In the episode The Sweet Smell of Success of The Raccoons, a Canadian animated series that ran through the 1980s, a band called the Raccoons had become the biggest act in Evergreen Forest. Bert played guitar and sang, Ralph was on drums, and Melissa on the keytar. When Cyril Sneer, the forest’s resident industrialist, found himself with an unsellable new cologne called Success on his hands, he decided the band was his way out of it. He brought a contract to all of them. Ralph and Melissa and the rest turned him down almost immediately, wary of anything with Cyril Sneer’s fingerprints on it. Bert did not turn him down. Wined and dined by Sneer and his lackeys, promised fame across the globe, a career, a lifetime supply of peanut butter, he signed.

What the contract actually bought him was a series of stunts he never agreed to and never wanted, culminating in a hot air balloon rescue gag gone wrong, Bert dangling from a branch over a cliff edge with nothing under him. Sneer and his men, watching it happen, simply left. No one came from Sneer’s side. Bert sang, alone and off key, that he should have listened to his friends, and it was the friends he had turned his back on, Ralph and Melissa and the rest, who showed up and pulled him back over the edge. The commercial never made him famous, and he barely appeared in it.

Remi Schmaltz was a farm kid outside Beiseker with three channels to choose from, and The Raccoons went out at 7:30 on a weeknight, early enough in the CBC’s primetime lineup that the whole family was usually still in the room. He watched it the way every other kid in the country watched it, week after week, without ever once thinking about the impression it left on him and how he would choose to do business later on in life.

Childhood

The Schmaltz name arrived in Canada by way of the Dakotas, and before that by way of Odessa, in the part of the Russian Empire that is now Ukraine. Late in the 19th century, the family settled just north of Beiseker, Alberta, on land that still demanded a sod hut before it would let anyone build anything more permanent. They farmed first, and the equipment that farming required became a business of its own within a generation.

The company was called Beiseker Motors, a Ford and Massey dealership moving tractors and short line equipment through the boom years of prairie mechanization. It passed from Remi’s great-grandfather to his grandfather and his grandfather’s brothers. His grandfather, a man everyone in the community called “Soup” Schmaltz, noticed something the equipment business alone could not give him. The dealership already had the relationships. Every farmer who bought a tractor from Soup Schmaltz was a farmer who would eventually need fertilizer, and so bagged, granular fertilizer became a side hustle run alongside it.

The equipment dealership was the business with the name over the door. The fertilizer trade started almost casually, to serve customers who were already there. Hard years in the equipment industry eventually closed Beiseker Motors for good. Soup Schmaltz carried the fertilizer arm forward on its own under a new name, Beiseker Agri Services, and ran it successfully as a dealer there for years.

Remi’s father, Soup’s son, left home after high school and disappeared from the family business for the better part of a decade, travelling through Europe and India in the 60s and 70s with no fixed itinerary and, for a stretch, no fixed address at all. He lived for a time in a treehouse in India.

When he finally reappeared, Remi’s father did not settle quietly back into Beiseker Agri Services. He took it over with ambitions that outran what Soup had built, determined to construct a large fertilizer blending plant of the kind he believed farmers would soon need as their operations scaled up. He took the business forward alongside his brother and brother-in law and expanded it from one location to four, adding chemical, seed, and custom application to the original fertilizer line. The company took the name Dynagra.

Remi’s mother’s side of the family came from the opposite direction entirely, east rather than west, and arrived three centuries before the Schmaltzes did. She was born and raised in Quebec City, and her ancestors were among the first French soldiers to arrive on the continent in the 1600s. Outside of First Nations lineage, he says, he might be as Canadian as it is possible to be. When she eventually made her way to Calgary for school, she arrived speaking only French. Learning English was part of the reason she came.

She became the stay-at-home mother of three sons. “Busy” is the word Remi uses to describe the job. Tasha came first, two years older than Remi. Justin came last, seven years younger.

Between Remi and Tasha there was closeness rather than rivalry, and Remi credits at least part of that to the fact that his older brother was, by then, a two-time national judo champion. There was not, he says jokingly, a great deal of fighting. He submitted to the will. The two of them ran in the same social circles, shared friends whose younger siblings became Remi’s own friends, and spent long stretches of adolescence together in an old shed on the property fitted with a pool table, hosting what Remi calls shaker parties that sometimes drew a couple hundred people. Risk was played. Cards were played. Pool was played more or less constantly.

Beiseker, when Remi was growing up there, had a population not far off from what it carries today, roughly 800 people. He went to Beiseker High, a school with roughly 100 students total, small enough that everyone knew everyone, and everyone’s parents, and by extension everyone’s business.

He was 13 when his father bought a snowmobile for the farm, one of the big sleds from a heavy snow year, and Remi decided he wanted a different one. He called a dealership in Calgary, found a salesman, and worked out a trade over the phone. The salesman had no idea he was talking to a kid. A few days later Remi and his father drove into the city to close the deal, and when Remi found the man on the floor, the salesman looked at him and said, “Sorry, were you the guy on the phone?” Remi said “Yeah, yeah, we had this deal.” The salesman looked at Remi’s father, then back at Remi, and said, “Oh my god!”

He was, by his own description, the kid who always won the door-to-door fundraiser sales, frozen food and whatever else the school was pushing that particular season, out on his bike, dirt bike or quad depending on the terrain, selling relentlessly and taking home the prizes for it. He liked the challenge of it and was never afraid to knock on a stranger’s door, so long as he did it, in his phrase, in a respectful way.

In high school, he rebuilt a 1954 Chevy from the ground up.

University

He enrolled at the University of Lethbridge in a general management program. He didn’t know what he wanted to do, so he picked something general, and marketing sounded interesting. He liked the idea of why people buy things.

In his third and fourth years he wrote business plans as class assignments. Instead of inventing hypothetical companies, he wrote plans for real initiatives inside Dynagra, the family business, plans that got implemented after he graduated and went on to generate millions of dollars in revenue. His professor gave those papers a D-.

He asked the professor if the plan was written poorly. “No, it was written fine,” the professor said. “So why the grade?” It turned out the professor knew nothing about the fertilizer industry and had never spoken to anyone in it. So how, Remi asked, was he even evaluating this.

Remi was never known for his academics, and he lost a lot of respect for the school system after that. What he got out of university was the relationships. To this day, some of his closest friends and best contacts are people he met there, people he has since built businesses and gone on adventures with.

It was also where he found the outdoors. Growing up he skied maybe five or ten days a season. At Lethbridge he joined the volunteer ski patrol at Castle Mountain, a couple hours away, compressing his classes into two days a week so he could live up at the patrol shack the other five, going out on avalanche control runs with the rest of the patrol. He calls it glorious. It was there that he also picked up whitewater canoeing and big game hunting.

The Family Enterprise

Life did not wait for him to finish deciding what he wanted to do with it. While he was still an undergraduate, his son Sage was born. He was engaged to marry Dominique, the woman who would become his wife. He had a mortgage and a minivan before he had a diploma. He described himself as a bit of an overachiever, shot out of a cannon, life arriving all at once rather than on any negotiated schedule.

His father had retired from Dynagra by then, leaving the business under interim management. “Just come work on some projects,” his father told him. He agreed, and found himself hooked almost immediately, not on any particular industry but on the underlying activity of business itself: identifying a real problem, understanding it, and building the thing that solves it.

His first real assignment was building out a finance division inside Dynagra to support the sale of crop inputs to farmers, work that put him directly under the mentorship of a retired executive from Farm Credit Canada who had built that organization’s own lending program from scratch.

Meanwhile, in a different corner of the same company, Tasha had begun experimenting with precision agriculture and variable rate technology, the practice of using field data to vary how much fertilizer or seed gets applied across different sections of the same field rather than blanketing it uniformly. The two brothers and Garth Donald, their head agronomist, gravitated toward that division together, hired people into it, and eventually filed patents on the technology. They realized that they were more passionate about the technology itself than the underlying business of selling crop inputs to farmers.

So in 2012 they sold the crop input business, the one their great-grandfather and grandfather had built, to CHS, the large American cooperative, and rolled the technology division, which had been operating under the name Dynagra VRT, into a new and independent company. They called it Decisive Farming. It is the moment Remi identifies as the real beginning of his startup life, even though nothing about how it started resembled the mythology of a garage-founded startup. They had nine or ten employees already in place, office space, and customers. Nobody in the room ever said out loud that they were doing a startup. They had a product they believed in, precise placement of fertilizer and seed using satellite imagery and soil ground-truthing, delivered through equipment the farmers already owned, and it already worked for the customers using it. The plan, insofar as there was one, was simply to keep going.

Decisive Farming

Keeping going turned out to be harder than the momentum of the spinout suggested. Growth was slower than they had projected, and it became clear almost immediately that the business would need outside capital in a way the family enterprise never had. Remi found himself at Randy Thompson‘s angel investor bootcamp in Calgary, absorbing an entire vocabulary of pitching, SAFE notes, and convertible notes that he had approached with essentially zero prior fluency. It felt, he says, like “drinking from a fire hose.”

The Canadian capital ecosystem of that period, roughly 2013 and 2014, did him few favours. The angel network and the family office community were, in his blunt assessment, weak and thinner than they are even today, but he ended up finding the lead investor for the company’s first meaningful round not in Calgary or Toronto but in the United States. Decisive Farming would go on to raise four rounds of capital in total, the first two angel-led and the latter two institutional, with McRock Capital leading the Series A in 2016, a relationship with McRock’s Whitney and Scott that Remi describes as still warm today. He is pointed about the exception to an otherwise clean cap table. In its seed rounds, the company raised capital exclusively from American investors, apart from friends and family. The only exception was a single Canadian angel investor who participated after the round was already closed and oversubscribed. He does not disguise his frustration with what that says about the domestic investment climate. Canadians, in his view, are simply too conservative with early capital, and it is an ecosystem-level problem rather than a personal grievance.

Building the technology itself required solving a gap neither brother could close on their own. Tasha was not, and is not, a software engineer, and Remi is candid that finding someone who could actually translate their domain knowledge into a functioning product took considerable trial and error, cycling through people before finally finding Marc Wachmann out of Calgary, the technologist who Remi credits with getting the business to where it needed to be from a purely software standpoint. The first customer, when it finally arrived, came from the same well the family had always drawn from. A farmer who had already been a customer of the crop input business was simply asked whether the new variable rate product made sense to him. “It did,” he said, “let’s try it.” The trust economy of Beiseker, again, doing its work at scale.

The near-death experiences, when they came, were less about product-market fit than about the sharp elbows of a small, well-funded competitive field. A well-capitalized rival threatened lawsuits outright, and Remi describes navigating a period of what he calls bullshit poker with them, both sides bluffing about legal exposure neither fully wanted to test. That competitor had acquired a patent on related technology and began telling conference audiences, in blunt verbal terms, that it intended to sue everyone in the space. Remi received a cease and desist letter from them directly. Shortly afterward Decisive’s patent was issued to them giving them room to manoeuvre, and Remi responded by sending a cease and desist of his own, an aggressive counter-move that led, improbably, to an invitation from the competitor’s investor, a large Silicon Valley fund, to fly down and sort the whole thing out in person. The meeting was informal enough that Remi describes it as shooting the breeze, and it ended with both sides agreeing, in essence, to mind their own business and leave each other alone. He draws a lesson from the episode that has stuck with him since. Patents in a startup context, in his view, function best as a defensive strategy rather than an offensive one, a shield rather than a weapon, and that the same aggressive competitor eventually had its patent squashed by another rival purely, as far as he could tell, out of spite. The industry, in other words, punished the very tactic it had tried to use against him.

Through all of this, Remi maintained a philosophy of staying deliberately friendly with competitors rather than treating every interaction as zero-sum, and it paid an unexpected dividend years later. At an industry event, he ran into Todd Ormann who at the time worked at that competing firm. Remi and Todd had a positive discussion as competitors. Years later Todd was the first agriculture employee at TELUS. Todd reached out to Remi to explore a potential partnership. His first instinct, by his own admission, was suspicion, the old worry that a big incumbent circling a smaller company is there to extract what it knows and then walk away. Todd did the opposite, over 6 months they had numerous meetings exploring a partnership opportunity that eventually led to Todd expressing interest to acquire Decisive Farming. It was that they respected the value Decisive Farming had built for farmers, and wanted to be part of extending it.

TELUS

TELUS’s acquisition approach came in March of 2019, over the phone, and Remi’s first response to it was no. He had, by then, developed a working framework for evaluating any offer against three simultaneous obligations:

(1) what it meant for shareholder value,

(2) what it meant for customers,

(3) and for its team.

A deal, in his view, only made sense if all three constituencies came out ahead together. An acquisition that served investors at the expense of customers, or vice versa, was not a deal worth doing regardless of the number attached to it.

It took roughly a month of back and forth before the conversation moved anywhere real, largely around whether a potential transaction could clear the return threshold his existing investors would actually require. Once that hurdle looked surmountable, TELUS put forward a letter of intent, and the following six weeks were consumed by negotiation over its terms. A deal appeared with a close date projected in August, but TELUS’s restructuring around hybrid tax treatment intervened, adding a further and, in Remi’s word, painful stretch of delay to a process that had already run almost a year.

It was during this stretch of prolonged limbo that Remi did something small and, in its way, perfectly characteristic. He had been clean-shaven his entire life up to that point. In early September, as the end of August deadline he had been promised slid by, he stopped shaving. Every additional week the deal dragged, his beard grew a little longer. People at the table started asking him about it. “What’s with the beard, Remi?” He would point to it and tell them “It’s taking you this long to get the deal done.”

The deal finally closed on December 12, 2019, the beard by then fully grown in. TELUS acquired Decisive Farming in a super successful exit being awarded as the most significant transaction that year, good for the team, good for customers, good for investors, the kind of outcome the three-part framework he had insisted on from the outset had actually managed to deliver.

He stayed on with TELUS’s strategy team for a couple of years after the acquisition closed, and it was the first time in his working life he had ever operated inside a genuinely large corporation. TELUS acquired 14 agriculture and food companies during his time there, a scale of integration complexity that had no analogue in anything he had built himself. He is generous about the culture he found there, and clear that he believed in the direction the company was trying to go. But he felt a mismatch almost immediately. He is, in his own repeated phrase, a “get stuff done” kind of guy, someone whose sense of purpose is bound up in visible momentum, and it became apparent fairly quickly inside a large corporation that individual momentum was not really how the work there was structured to happen. He did not experience this as a betrayal or a failure on anyone’s part. He experienced it as information and began quietly planning an exit of his own.

Before that plan could fully mature, the accumulated toll of the preceding years caught up with him. He had burned out once already, in 2017, in the run-up to the Series A, a hole he says took him roughly two years to climb fully out of. The TELUS acquisition process and its aftermath dug a second one. He and his family took ten days over Christmas in Belize to go sailing and to disconnect.

Whitewater Trip

The world outside his own recovery was simultaneously navigating the early months of the COVID pandemic, which he notes made an already disorienting period feel stranger still. He spent much of that stretch mentoring other founders through Thrive and Creative Destruction Lab, work that gave him proximity to other people’s problems without requiring him to solve any of his own, and it was in the course of that mentoring, watching other entrepreneurs raise capital and structure deals, that he arrived at a conclusion about what he personally still needed to learn.

If he was ever going to start another company, he needed to understand the investment side of the business first. Without that, and especially as a Canadian founder, you are hamstrung relative to your American counterparts. So he made a deliberate effort to network inside the investment community itself, to learn the mechanics from the inside rather than across the table.

The result was Koan Capital, launched with a $5 million fund that got raised in seven weeks, an unusually fast fundraise built around a pre-seed and seed stage agtech thesis, funded entirely by non-institutional capital, and currently deployed over 4 years across roughly 18 companies spanning Canada and the United States. He describes the experience as extremely rewarding on terms that have little to do with the eventual returns. He learned, and continues to learn, from the entrepreneurs and co-investors he works alongside, in a way that fed directly and consciously into what he built next. Brilliant Harvest, when it finally arrived, did not spring from nowhere. It came out of a period of study he had specifically engineered for himself, and by his own account it started, quite literally, on a napkin.

The napkin sketch found its final shape on a whitewater trip in Idaho in the summer of 2023.

He was on the South Fork of the Salmon River with his son Sage and roughly six other people, running a multi-day Class V stretch he had been building toward for years in his own paddling career, in the twelve-foot cataraft that had become his boat of choice. His oar caught a rock in shallow water, shot upward, and struck him squarely between the eyes. He had never had a concussion before, and on a multi-day river trip far from any hospital, the only person who can reliably diagnose one in the field is the person who has it. His group asked if he was okay, but nobody actually said the word ‘concussion,’ so he kept paddling.

The following day, on a gentler stretch of the main Salmon, his guard was fully down. He had lent his boat to one of the other kids, and while joking around near his son’s kayak, he tried to flip it as a prank. His son reflexively swung the paddle back and caught him in the side of the head. What followed was, in Remi’s own words, a series of bad decisions. He got into a kayak he had no real business paddling, since he does not have a roll, swam once, and was approaching the biggest rapid on the stretch when one of his son’s friends, a genuinely excellent kayaker who happened to be twenty feet ahead of him, looked back and told him, “I’ve got you, Remi.” They made the line. It was not until breakfast afterward, at an old-fashioned diner with pen-and-paper ordering, that the injury actually announced itself. The waitress asked what he had ordered. He had no idea. That was the moment he knew.

The recovery period that followed, weeks of no screens nor reading, is where Brilliant Harvest was actually born. Remi is unapologetic about how loosely he interpreted his doctor’s instructions. He was cleared to do anything that did not hurt, and he took that as licence to keep talking to people even while he stayed off screens entirely. In the quiet, he ran a private diagnostic on himself, three questions he kept coming back to.

What am I good at? What do I know? And where is the money — meaning specifically where is the money attached to an unsolved problem?

He knew agriculture and was excellent at sales. What he needed was the third piece, the actual problem, and it arrived from a familiar source. Doug Heritage, a Manitoba farmer who had been a Decisive Farming customer years earlier and had since become a genuine friend, was complaining to him about how painful dealership service had become, how expensive and scarce good technicians were, and how a farmer could spend well over a million dollars on a single piece of equipment and still struggle to get a straight, timely answer to a basic operating question.

The timeline that followed moved fast for a man supposedly convalescing from two concussions in two days. The second injury happened on July 5th. By July 15th he had validated the problem, talking to farmers, dealer experts, and people across the equipment industry. By the end of the month, Mark Blackwell at Builders VC, who had spent a few months sharing ideas with Remi that never panned out, agreed this one did fit, and offered to write the first cheque. The company was incorporated on August 1, 2023.

Brilliant Harvest

Brilliant Harvest is built to solve a specific and, once explained, obviously enormous problem. An equipment dealership might carry and support 20 to 100 different brands, an amount of technical knowledge no individual employee could realistically hold in their head, compounded by manuals that can run to 12,000 pages for a single model. Over 50% of the questions a farmer brings to a dealer are repeats, or easily answerable, but during planting and harvest, when every farmer in the region needs help at the same moment, even the easy questions become bottlenecked behind busy staff, and staff themselves often have to search across half a dozen internal systems just to locate the answer.

Brilliant Harvest’s platform ingests a dealer’s full library of equipment manuals, connects directly into the dealer’s business systems to pull work order history, parts usage, and customer records, and processes conversational data from call centres and internal technician group chats where informal best practices already circulate. Out of all of that it builds a proprietary, dealer-specific knowledge base, an asset that belongs to that dealer.

The near-crisis Remi identified in the company’s first two years was not financial or competitive but philosophical, a question of restraint. Everyone in the industry, he says, was talking about AI, and very few were implementing it in any way that was actually profitable or sustainable as a real business. The temptation he had to resist, repeatedly, was over-engineering the product with AI where a simpler interface would have served the customer better. Humans, he points out, are pretty capable on their own. Give someone a couple of buttons to choose from and they will pick correctly almost every time, in situations where asking an AI system to correctly infer the same intent from open text is a genuinely harder problem to solve reliably. Given the stakes involved, heavy farm equipment where an unreliable answer carries physical risk rather than mere inconvenience, the company has erred deliberately toward conservatism in what it deploys to production.

Part of why he trusts the team to make that judgment correctly is who is on it. Five members of the founding team had worked with him previously at Decisive Farming, and the leadership team is proven with deep domain expertise, giving the company an unusual head start on internal trust and communication. The Head of AI, Micheal Lanham, had already authored roughly a dozen books on applied AI development, including two specifically on deploying AI agents, and Remi still seems pleased, almost surprised, that he found someone with that particular depth of expertise sitting in Calgary.

Brilliant Harvest has grown to 34 employees, serving more than 300 equipment dealer locations across Canada and the United States, representing over 70% of CNH and more than 30% of AGCO’s large dealer networks. Most of those dealers sell and support both agricultural and construction equipment, which has opened a second, adjacent market for the same underlying platform.

Convictions

Two convictions sit at the centre of how Remi runs a company, and they appear, on the surface, to pull in opposite directions.

The first is that mistakes are not just tolerated but structurally necessary. Nobody, in his account, has ever been fired across any of his companies for making a mistake outright. People get fired when they repeat the same mistake, when the lesson visibly fails to land, or when a mistake gets hidden rather than surfaced. Founders and their teams are, by definition, operating in territory nobody has fully mapped before, and a team that is not making mistakes with some regularity is a team that is not actually pushing hard enough.

The second conviction is that he focuses on culture, independent of performance. He does not take hiring or firing lightly, he says, but once it becomes clear someone genuinely does not fit the team, he moves. What has surprised him, every time, is the reaction from the rest of the team afterward. People consistently come back to thank him for making the call, which he takes as confirmation of a broader principle he holds about what actually destroys startups. Large corporations can survive, even if imperfectly, when individual employees are primarily optimizing for their own career advancement rather than the shared mission. Startups cannot. That misalignment is how startups fail, full stop, and it is why he treats cultural fit with the same seriousness other founders reserve for product-market fit.

Underneath both convictions sits something older than either of them, an old school mentality he traces back to the farm, around your word, your honour, and accountability. In a small community, where everyone knows everyone and everyone knows everyone’s parents, there is a deeper trust already embedded in every relationship before a single transaction has taken place. Remi had recently signed a new customer, a dealer who told him he had made three reference calls before committing. The dealer’s verdict was not really about the product. It was that Remi and his team were accountable, present. “Stuff breaks,” the dealer told him, “that is just the evolution of leading technology, but you are there, you are fixing it, listening and accountable.” “That,” the dealer said, “is who we want to be working with.”

Remi might not remember the exact lesson of The Raccoons’ Sweet Smell of Success episode. Cyril Sneer offered Bert a contract, promises, and peanut butter, yet placed him in danger and left him in the trees. Ralph and Melissa showed up to save Bert, even though they didn’t need to, but because it was the right thing to do. Remi would tell you the real-life lessons of the cartoon episode came later, on the farm outside Beiseker. Remi spent the next three decades making sure he always showed up when it mattered.

The word he returns to more often than any other across every conversation about his career is symbiotic. A healthy company, in his framing, requires a genuinely reciprocal relationship between the people inside it and the company as an entity, each one actually needing the other to thrive rather than one simply extracting value from the other. It is the same word he uses to describe the relationship Brilliant Harvest is trying to preserve between manufacturer, dealer, and farmer.

Remi does yoga three or four times a week now, holds a brief morning meditation practice, nothing elaborate, and has become methodical in a way he was not during the Decisive Farming years, when he is candid that work-life balance was genuinely poor and his wife Dominique absorbed a real share of the cost for it.

Dominique is a Banff native born into adventure in the mountains, and is a sales consultant for a high end art gallery. Living in Banff is a lifestyle Remi describes with evident fondness, small-community and unhurried, the kind of pace he has come to value more deliberately with age. The couple has raised three children whose lives orbit the same outdoor obsessions that first pulled Remi in during his university years. Sage, the oldest, is a serious whitewater kayaker, skier and fisherman, home in the summers from university as a fishing guide. Middle daughter Olive; the witty one; is in 2nd year civil engineering and is an avid hiker, skier and loves books. The youngest Hazel, known for her forward candid demeanor allows her to connect with all ages and demographics, it’s truly a super power. She is a great whitewater kayaker and skier. It’s a family that bonds over the outdoors itself, which functions, in practice, as the family’s actual shared language.

The future of artificial intelligence looked bleak.

The field had spent the 1950s and 1960s promising machines that would translate languages and reason like people, but had failed to deliver any of it, and had watched governments on both sides of the Atlantic withdraw their funding through the 1970s in what the field itself would eventually name an “AI winter.”

What survived was expert systems, which worked by having engineers interview specialists and write down their knowledge as thousands of hand-written rules, and which were enjoying a brief commercial boom that would collapse within a decade. Neural networks, the idea of building something that learned from examples rather than being told the rules, had been effectively dead since the late 1960s, when an influential book demonstrated the limitations of the simple versions and the money went elsewhere. The people still working on them could be counted by hand, and most of them had been advised to work on something else.

By the 1990s, the words “artificial intelligence” had become slightly embarrassing to use. Researchers who wanted funding called what they did machine learning and borrowed their methods and vocabulary from statistics, and they talked about specific narrow tasks rather than about intelligence, which was what the previous generation had promised and failed to deliver.

None of this reached a boy in a town outside Montreal, for whom the words still meant whatever he had seen in video games.

Childhood

Hugo’s parents were both Québécois, descended from settlers who came over from France generations back. His mother stayed home through most of the years he was in school and later returned to the education system as an éducatrice spécialisée, working alongside children with learning disabilities and other difficulties, supporting them in the classroom. His father was a mécanicien d’entretien, a maintenance mechanic, the person a factory calls when a machine breaks or the line stops working. Nobody in the family had been to university.

Mathematics was Hugo’s favourite subject at school. He was curious about computers from a distance, since the family did not own one. He worked on convincing his father, insisting that they should buy a computer, until his father gave in and one came into the house. He then barely programmed on it at all.

His programming life began from a calculator. Graphing calculators had started appearing in Quebec high school mathematics classes, and teachers spent time showing students how to plot functions and solve equations with them. Buried in the manual was a programming feature, a simple language that let you write instructions the calculator would carry out in sequence. Around his fourth year of secondary school he decided to find out what he could build with it, so he built a map drawn out of text characters, with a figure standing on it who could be moved from one tile to the next, where arriving at certain tiles meant finding a treasure to pick up. He wrote a small puzzle story around the movement. It was the first thing he ever coded, it took him roughly a week, and he was proud enough of it to show his friends and several of his teachers.

There was, however, nowhere at school to take it further. He wanted to enroll at CEGEP in the computing program and pushed his parents hard on it. They pushed back, having been told by a teacher during a school visit that a student strong in the sciences would be bored inside a technical computing diploma, so he went into pure sciences instead. It annoyed him at the time. The computing electives on offer at CEGEP taught Word and Excel, and the only real programming course he took before university was a single class. But, his parents turned out to be right. The mathematics he took instead included content that was nearly impossible to learn later, and the programming was not.

Prior to CEGEP and during, Hugo had spent years watching American sitcoms to learn English. The Fresh Prince of Bel-Air was the first one he watched, and he could never say why afterwards, only that something in the way they spoke made it possible for him to understand. He followed one episode, then followed them without effort, and some time later he passed an episode of Home Improvement, which he had failed to follow before. He watched an episode and found he could understand all of it.

The show he loved most was Whose Line Is It Anyway?, a 1998 American improv programme airing on ABC in which performers were given a suggestion and had to build a scene out of it on the spot, and he watched it more or less alone, because it was an anglophone show and nobody he knew had heard of it. Then he mentioned it to a girl at CEGEP. She told him she loved it too and that she had never met anyone else in Quebec who did. On their first date they went to the cinema, got the showtime wrong, and missed the screening entirely, so she came back to his house (which was his parents’ place as he was still living with them), and they watched his recordings of the show instead. He ended up marrying her years later.

In his second year of CEGEP he entered a mathematics competition open to colleges across Quebec. He did not win, though. He finished somewhere around 20th or 30th, which was not high enough to mean anything, but high enough to clear the bar for a summer school at the Université du Québec à Montréal, where professors brought in from several universities taught the students who had placed well more advanced material than any of them had seen.

Before that summer began he had to choose a university. McGill and the Université de Montréal both accepted him, and McGill offered the better package, a scholarship renewable every year against a single year of funding at Montréal. Anyone asked which of the two carried more weight internationally would have said McGill without hesitation. However, he turned it down. He was not confident his English was strong enough to survive four years of coursework in it, and he was afraid he would struggle, so he enrolled at the Université de Montréal in computer science.

Then he went to summer school. One day, he sat down to lunch with a professor from the Université de Montréal, and she asked what he intended to study. He said computer science, and told her he was interested in artificial intelligence, a term he could not fully define, having met it mostly through video games and found it intriguing without knowing what it entailed. She told him there was a great deal of mathematics inside artificial intelligence, and that the joint mathematics and computer science degree would serve him better than the one he had signed up for.

He had weeks left before the term began. He changed his degree on the strength of one conversation across a cafeteria table, with a woman he had met that week and would never study under. It was an impulsive decision, and out of character, and he has never been able to explain why he made it at that time.

Université de Montréal

In his first year, a professor teaching one of the programming classes noticed his grades and told him about a federal program that funded undergraduates to spend a summer inside an academic lab. The pay was roughly $6,000 for three months, which was not a fortune but better than his job at McDonald’s in every possible way. He applied for the program, got accepted, and quit his job flipping burgers. The internship was in RALI, the natural language processing lab at the Université de Montréal, under Philippe Langlais, and it was his first exposure to a research problem.

However, it was not where he originally wanted to be. He had looked up which professors at the university worked deeply on AI, and one name came back in the search results. He emailed Yoshua Bengio to ask whether he was taking summer interns. Bengio replied that he did not have enough mathematics yet and that it was too early for him.

Some time later Hugo went to Bengio’s office to ask which classes he ought to take, and Bengio listed them. Almost all of them were statistics. Introduction to statistics first, then linear regression, then sampling, then a course on statistical estimation that spent the term on the properties of estimators, then statistical testing. Hugo walked out depressed, holding a list of the courses he would least have chosen for himself, because statistics was the part of mathematics that he enjoyed the least. The first course handed out formulas without explaining where they came from, which made the whole subject feel arbitrary and engineered rather than derived. He took them anyway. By the second course the probability arguments behind the formulas started showing up, and by the time he reached linear regression he was enjoying himself, having worked out that he was looking at a very simple form of machine learning and that you could say mathematically grounded things about how a predictive model behaves.

He came back the following summer, and Bengio took him under his wing. Another student arrived at the same time, a friend from the computer science program who had not taken mathematics, and who gave up after about three weeks and found an internship elsewhere. This was the summer Hugo first learned what a neural network was and what backpropagation did. He found it hard, but made it through, and he is certain he would not have without the dual degree he switched into.

The lab he had walked into was working against the consensus of the field. The methods that dominated machine learning then were kernel methods, methods associated not with non-convex optimization but convex optimization. This was considered a big breakthrough, being able to build a nonlinear model and frame it as a convex problem, because then you didn’t have to deal with difficult optimization problems. The methods also came with convenient mathematical properties that neural networks did not have, and a body of theory characterizing how well they would generalize. Their limitation was that they generalized locally. A kernel method handles a new photograph by leaning on the training photographs nearest to it, which works as long as the new thing sits close to something already in the collection. Show it a mango when it has only ever seen apples and oranges, and it will reach for whichever of those looks closest and call it that, with no way of registering that it is looking at a different fruit altogether.

Bengio believed that the biggest challenge in the field was finding methods that could generalize further from the training data than local kernel methods could manage, and that this was what the lab was trying to tackle from the beginning. His answer was the neural network, which learns a general account of the data rather than storing examples and comparing against them, and which might therefore say something sensible about a mango. Very few researchers anywhere were still working on them. They were regarded as a black box that took a kind of dark magic to make work, and success with them was thought to depend on unreliable insider knowledge passed hand to hand.

Bengio ran his lab as an open floor. If he had an idea he said it to everybody at once and hoped somebody would pick it up and run with it, which gave the room a strong sense that everyone was working on the same problem, and occasionally produced the discovery that two people had been chasing the same idea for a month. Hugo stayed there through a master’s, then a direct transfer into the doctorate on a scholarship, then the doctorate itself. He never seriously considered leaving. He had caught the bug, and there was too much left to learn.

The other thing he learned in that office had nothing to do with mathematics. He was sitting with Bengio one afternoon when the phone rang, and Bengio picked it up and began speaking French in a Parisian accent, which Hugo had never heard him do, having only ever known him with a Quebec accent. He was thrown by it. For a Quebecer, the French accent carries a whiff of pretension, and he sat there wondering who was on the other end of the line and why they warranted a different voice. He learned later that Bengio had been born in France and partly raised there before doing his schooling in Quebec, and that the switching was not performative but reflex, tuned to whoever had spoken last. In a three-way conversation with Nicolas Le Roux, a French classmate doing his PhD alongside him, Hugo would watch Bengio flip accents back and forth across a single conversation depending on which of them had opened his mouth. At the time it struck him mostly as evidence of how small his own world was. He had never been outside Canada, nor had ever been on an airplane. The first flight he ever took was to a machine learning conference, and he was nervous about it.

NIPS Conference

The conference was NIPS, around 2004, and he went with no paper and nothing to present, purely to read posters and find out what the field looked like.

Almost nothing there was about neural networks. He remembers something like four posters on the subject, off in a corner, and the rest of the hall given over to kernel methods, support vector machines, nonparametric Bayesian models, and a great volume of theoretical work proving generalization bounds. He could not read most of it. Bengio had taught his students neural networks and comparatively little of the learning theory everybody else in that room had grown up on, so Hugo spent the week walking past work he did not recognize, written in a language he had not learned, all of it more fashionable than his own.

“Am I studying the right thing?” he remembers thinking. “That doesn’t seem to be what other labs are studying.”

He was studying the right thing, and it would take the better part of a decade to prove that. In the meantime, the fact that so few people were doing it meant his expertise was worth far more when the field finally turned.

His work in those years went after a specific weakness in the kernel methods. The task was density estimation, modelling the distribution of some dataset of inputs, which at the time was still largely unsolved. The kernel approach placed all training examples and assumed a small Gaussian around every training example and assigned equal probabilities in every direction, quickly decaying as you got away from the training example, and taking the average of all these Gaussians, like a huge non-parametric Gaussian mixture. The problem was that a model given a pile of images had no way of working out which variations on them counted as real and which did not. A handwritten digit can be rotated, shifted in the frame, or drawn with a thicker stroke and remain the same digit, but the ways it can change are tiny in number next to the ways its pixels could be rearranged.

Bengio’s earlier answer was to work out those permitted variations separately around each example, using its nearest neighbours. Hugo’s idea was that the structure had regularity to it, since any digit can be rotated and any digit can be shifted, so a common structure ought to sit underneath all of it, and one neural network could produce those directions across the whole space rather than rediscovering them example by example. The paper, Non-Local Manifold Parzen Windows, came out at NIPS in 2005. Its central experiment taught the model about rotation using nine classes of digits, then handed it a digit 1, which it had never seen rotated, and asked it to turn 1. It could, even rotating in the opposite direction from anything in its training data, while the local method it was measured against could not. The paper was part of what Bengio’s lab was doing more generally, which was to take the popular kernel methods, find the problems they solved badly, and show that a neural network solved them better.

PhD

Nobody was training with GPUs in 2006, so the field was looking for algorithmic contributions that would help it train larger networks with less compute. Geoffrey Hinton had proposed one. Rather than training a deep network all at once, which was hard to do, you pre-trained each layer individually, one layer after another, using the learning algorithm of a restricted Boltzmann machine.

Bengio’s lab was exploring the idea in many different directions, and Hugo’s contribution was to suggest replacing the restricted Boltzmann machine with something simpler, for a reason he states without embarrassment, which is that he did not understand restricted Boltzmann machines very well at the time and was looking for a way around them. What came to mind was a single layer that takes an image, passes it forward into a hidden layer, passes it forward again into an output layer the same size as the input, compares the reconstruction against the original, and adjusts itself according to the difference. For instance, if you give a photograph of an apple and ask it to reproduce that apple, and it succeeds then it has learned something about how apples are put together.

He did not know at the time that the idea had been explored before, under names like “auto-associator,” or that “autoencoder” would become the term everyone used. It did not work as well as the restricted Boltzmann machine, but it worked considerably better than doing no pre-training at all, and it was one of the first things he felt he had contributed from his own ideas that seemed to have legs. It went into the paper, Greedy Layer-Wise Training of Deep Networks, published at NIPS in 2006.

His version had a problem. If the hidden layer was as big as the input layer, nothing stopped the network from learning a representation that was simply a copy of the input, taking each individual pixel into its own hidden unit, copying it out again, and solving the problem trivially. Pascal Vincent was a junior professor in the department who had done his own PhD with Bengio, and Hugo happened to go into his office one day to talk about the work he was doing. Vincent suggested that a way of avoiding the copying solution would be to add noise to the input, so that the problem the network had to solve was not to reconstruct the original input but to uncover some of the noise that had been added to it. Hugo thought immediately that it made intuitive sense, and that a network doing that would probably need to learn something about the regularities in the input distribution and about what makes different pixels depend on one another.

The first time they tried it, it just worked. What they would do at the time was train on handwritten digits and then look at the filters learned by the hidden units, and what you were hoping to see, which is something you would see with restricted Boltzmann machines, were stroke detectors, filters that looked like a pen stroke with empty background around it. They were not getting that with plain autoencoders, which they thought was because the model was too tempted to copy each individual pixel. As soon as they added a bit of noise, the stroke detectors started to appear. The work went to ICML in 2008 as Extracting and Composing Robust Features with Denoising Autoencoders. Hugo is careful to say that at the start neither of them really knew why it worked, that the more formal account of how the method relates to the local structure of the data distribution was shown afterwards, mostly by other people in Bengio’s lab.

The other paper from those years is the one he regrets, though not for its contents. Hugo, Bengio, and a fellow student had built a system that could produce a classifier for a task it had never seen, working from a description of the task rather than from labelled examples of it, and they called this Zero-data learning. Almost immediately afterwards somebody else published nearly the same idea under the name “zero-shot learning,” which is the name the entire field uses today. For years afterwards people kept asking Hugo what zero-data learning could possibly mean, since learning from no data is a contradiction.

NIPS rejected the zero data-learning paper. Hugo sent it instead to AAAI, whose deadline fell a week before ICML’s, which suited him because it cleared one paper out of the way before he pushed the denoising autoencoder to ICML. The venue change also let him write a sentence he could not have written for a machine learning audience. In the mid-2000s the term artificial intelligence was very nearly unsayable in the rooms Hugo worked in, contaminated by expert systems and the failures of the previous generation, and the machine learning community had responded by borrowing heavily from statistics, narrowing its ambitions, and talking about tasks rather than about intelligence. AAAI had the words in its name. So in the conclusion, he and his co-authors wrote about extending machine learning towards AI, and described a system that would understand instructions specifying a task and then perform that task with no additional supervised learning, provided it had already been trained on related tasks described in the same language.

He was describing prompting, but he would not see it work for another fifteen years.

Toronto

He reached the end of his PhD without a plan.

He knew he liked research and wanted to stay in Quebec, and staying in Quebec meant academia, because there was no industrial lab in the country doing the work he cared about. No Google AI lab, no DeepMind, nothing. So he told Bengio he would do a postdoc and asked whether he should write to Hinton. The two were not collaborators but they spoke regularly, and the arrangement came together almost immediately, much faster than these things normally do, which Hugo puts down to a phone call in which Bengio presumably said something generous. He applied for a fellowship, won it, and moved to Toronto.

He was a nervous traveller by his own description, and living in English made him nervous even inside his own country. He had a daughter by then, born during the PhD, and a second would arrive while they were in Toronto, and he did not want to be far from his parents, who would be needed. The United States was the obvious move for someone in his position, but he did not seriously consider it for those reasons.

Hinton ran a lab that was the inverse of Bengio’s. Where Bengio broadcast ideas into a room and let them find an owner, Hinton gave each student and postdoc a project that was theirs from the first day and understood by everyone to be theirs, which Hugo found clarifying. His fellowship was generous enough that his salary was not coming out of Hinton’s grants, which bought him room to work on things that were not Hinton’s projects. He used it to join Rich Zemel’s reading group and finally learn the probabilistic graphical models that Bengio’s lab had never taught him.

The problem he picked was one that had frustrated him. There was no reliable way of evaluating an unsupervised model. A lot of the methods that existed at the time were looking at the filters a model had learned and forming an indirect judgment about the features. An apple is a useful way to see what was missing. A model that has learned what apples look like ought to rate a new photograph of an apple as ordinary and a photograph of scrambled pixels as nearly impossible, and that rating is the number nobody could get out of these models directly. Ruslan Salakhutdinov had started scoring the quality of these models by asking how much probability they assigned to real images they had never seen, which was the right idea, except that computing it involved an approximation, and the approximation was what frustrated Hugo. Working with Iain Murray, another postdoc in Hinton’s lab, he found a way to build a model that produced the number exactly. Rather than having the model describe a whole image at once, they had it describe one pixel at a time, each pixel predicted from the pixels before it, which meant the total could simply be multiplied out. The result was NADE, The Neural Autoregressive Distribution Estimator. It was the first paper he published with no supervisor on it, made by two postdocs in a room, and the impact it had was theirs.

He also spent time on an idea of Hinton’s about foveation, building a model that chose where to look inside an image, took a glimpse at that spot, accumulated what it had seen, then used the accumulated glimpses to decide what the image contained. Everything in that lab was an attempt to train bigger networks on more data through cleverness rather than force.

Hugo had left Toronto by the time AlexNet won the ImageNet competition, realizing how important it was to the field despite never sharing a conviction on it. Ilya Sutskever had become persuaded that the answer was to train a much larger model on a much larger collection of images and solve whatever engineering problems stood in the way, which meant abandoning the belief that a deep network had to be assembled layer by layer before it could be trained at all. Hugo had spent years helping to build that belief. The layer-by-layer methods turned out to be unnecessary. What the problem had actually needed was a better training procedure and a graphics card fast enough to run it.

YouTube Channel

Hugo returned to Montreal and took a faculty position at the Université de Sherbrooke, and discovered on arrival that he had never taught anything to anybody.

This is ordinary for new professors and nobody warns them about it. The first thing he learned, which he says almost everyone learns in their first semester, is that what is obvious to you is not obvious to the students, and that the correction feels like dumbing the material down when it is really just putting it at a reasonable pace. He was lucky in what he was asked to teach. They gave him introduction to artificial intelligence, then let him build courses on neural networks and machine learning from scratch, and never made him teach introductory programming.

Open online courses were becoming a phenomenon at the same moment, and Hugo looked at them and recognized himself as a student. He would have loved it, being able to watch a lecture at his own speed, stop it, watch a section again, sit with something difficult until it opened, then bring a question to a professor already knowing where his confusion was. So he inverted his classes. The lectures went onto video and the students watched them at home, and the class hours were given over to exercises, with him in the room while they worked, which is the arrangement known as the flipped classroom and which Sherbrooke’s department let him try. The results were at least as good as what he had been getting the conventional way.

The videos went onto YouTube, where they were watched by a great many people who were not his students, and this had an effect he had not anticipated. People began recognizing him at conferences and asking to take photographs with him. He had become a known name in deep learning partly through his papers and partly through a course he had recorded alone in front of a laptop.

Recording was strange at first, sitting in a room talking to nobody, and then he began to enjoy it, including the editing. Hugo walked into every lecture that term more prepared than he had ever been in his life, because a stumble would be visible forever and painful to edit and might mean shooting the whole thing again. He says he was never especially good at teaching, that he put himself in conditions where he was energized to do it, and that innovating on the format made it feel a little like research.

Hugo does not teach any more. He still answers questions under the videos, which scratches the itch, and he still gets comments telling him his French accent is strange, which he says he has learned to go beyond at this point.

Whetlab

During the Toronto postdoc he met Ryan Adams and Jasper Snoek, and together they set out to automate one of the most tedious jobs in the field. Every machine learning model has hyperparameters, settings like the learning rate and the number of units, which have to be chosen before training starts. At the time, people saw the training of a neural network as a black box and the researcher picked those values by hand. Their method used Bayesian optimization to choose them automatically instead, and the experiment that struck people was one in which it tuned the hyperparameters of a deep neural network better than the original settings the researcher had arrived at by hand.

The paper, Practical Bayesian Optimization of Machine Learning Algorithms, was finished after Hugo had moved to Sherbrooke, which makes it technically his first Sherbrooke paper despite having begun in Toronto, and it went to NIPS and very nearly did not get in. The reviews ran from lukewarm to negative. One said the work was not particularly creative, on the grounds that the underlying technique already existed and the authors had merely pointed it at a new target. The area chair disagreed, championed the paper against its own reviews, and pushed it through.

The machine learning community then decided it was one of the most useful things published that year. It became one of Hugo’s most cited papers, and at that NIPS it finished as the second most cited paper of the entire conference. The most cited paper that year was the ImageNet paper out of Hinton’s lab.

Interest followed, particularly from the large technology companies, which wanted anyone who could do this kind of work. Ryan Adams looked at the situation and concluded they should start a company. He convinced Hugo and Jasper Snoek, recruited Kevin Swersky, a Toronto PhD student working on the same problem, and brought in Alex Wiltschko, a neuroscientist who had met Adams and could actually build things.

Hugo was a part-time founder. He was still a professor with a lab and a teaching load, and he says his contributions were minor and the credit belongs to the others, who pushed hard enough to get a real product working, a service customers could query to receive tuning recommendations. He gave some input on product design and found the company its first client, a deal that never closed, because the company was acquired before the contract was signed.

The acquisition was of its time. There was so little deep learning talent in the world that the large companies were buying teams as readily as technology, and Twitter, which had people inside it arguing that it needed a deep learning laboratory of its own the way Google and Facebook had theirs, bought them. Hugo believes the technology was used, and also that the purchase was partly a signal, a way of announcing that Twitter took this seriously, helped along by the fact that several of them came out of Toronto.

Jack Dorsey returned as CEO in the middle of the negotiation. The founders panicked for a while about whether a change at the top would kill the deal. It did not, and Hugo met Dorsey a handful of times afterwards.

He took the offer for two reasons. The first was financial. The second had been bothering him for years. He had never worked in industry, never done an industrial internship, never built anything that had to be deployed, and he was training students who almost all went into industry, in a field that exists to solve concrete problems rather than abstract ones. He felt strange about not having that experience himself. Sherbrooke gave him a leave of absence, let him keep his affiliation and finish supervising his students, and he moved his family to Boston.

Game of Thrones

The research he did at Twitter came out of watching engineers do something tedious.

Engineers designed classifiers of tweets by hand, and they had to design a lot of them, because that was one way to track specific types of content on the platform or surface certain signals. Somebody would ask whether they could get a classifier about Game of Thrones, because the product team wanted to surface tweets about it when the latest episode had just aired. Hugo looked at that and saw a value in designing methods that could produce classifiers automatically from very few examples of the kind of tweets you wanted.

So he started on few-shot learning, which means building systems that can generalize from a handful of examples, using a family of methods called meta-learning, where you design learning algorithms that discover other learning algorithms. It was related to the hyperparameter work in Toronto, since both were attempts to automate part of a machine learning scientist’s job, except that this time the target was not the settings but the learning rules themselves.

The paper he published from Twitter, Optimization as a Model for Few-Shot Learning, contained no tweets at all, and tweets were its entire motivation. He is prouder of it than of almost anything else he has written, partly because of where it took the field, and partly because, years later, the Google Vice-President Blaise Agüera y Arcas gave a keynote at NeurIPS 2019 and said the work had inspired his own.

His second assignment at Twitter was to build a research culture inside a company that did not have one, and this is the part that failed. Twitter had not turned a profit in years and there was open speculation in the market about whether it ever would, and eventually the decision came down that there would be no separate research laboratory, that it would be folded into the product organization. Hugo realized that whatever had been brought in to provide was no longer wanted.

The timing was almost absurd. In the same period, Samy Bengio told him that Google wanted to open a Brain laboratory in Montreal, the first Google Brain team anywhere outside the United States, and was looking for someone with the experience and the appetite to build a team from nothing.

Google Brain

Google chose Montreal because of Montreal, which is to say because of Bengio and the ecosystem that had accumulated around him, and because reinforcement learning was becoming commercially important at exactly the moment when Doina Precup and Joelle Pineau were both working in the city. Hugo’s reason was that his family would be home.

There was already a Google office, so the office setup had already existed and he would not need to spend a year negotiating for desks. What he was given from the get-go was headcount. His first recruits were Marc Bellemare, Nicolas Le Roux, and Danny Tarlow, and two of the projects he is proudest of from that decade came from them.

Bellemare took on Loon, a Google moonshot that flew balloons through the stratosphere to deliver internet connectivity, and replaced the balloons’ navigation system with one that had taught itself to steer. Learning by trial and error was widely regarded then as a technique that worked beautifully in simulation and fell apart on contact with the physical world, and here was a fleet of real balloons crossing the globe under a learned controller. The work went into Nature.

Tarlow wanted to apply deep learning to code, which in the late 2010s was a niche interest almost nobody was pursuing. The first thing his group built was a system that suggested fixes when a build broke, trained on Google’s own record of what its engineers had actually done in that situation over the years, on the reasoning that a task performed constantly and captured in data is a task a model can learn. Tarlow is still at Google and is now central to Gemini’s coding abilities, and Hugo says with some satisfaction that his team was years ahead on what turned out to be the most popular use of AI in the world, today.

His own role had changed shape, and he has two ways of describing it. When he was a PhD student writing papers he owned, he was an artist making albums, and he had since become a producer, making other artists’ records. The other description is gardening. The researchers around him had doctorates and reputations and did not need direction, and telling them what to work on was neither possible nor desirable, so the job was to prepare the conditions in which good research grows, meaning the physical space, the ease of collaboration, and the chance encounters between people from different fields.

Through all of it he kept one day a week at Mila, where he held an adjunct professorship and supervised students, some jointly with professors there. He had been trying to knit the Montreal ecosystem together since he was a graduate student, when he and Doina Precup started alternating their two labs’ seminars between the Université de Montréal and McGill, for no reason beyond its being absurd that two groups working on the same problems in the same city should sit separately. Mila became a national institute in 2017, alongside Vector and Amii. Hugo’s view of his own position was that it mattered not only for Google to be in Montreal, but for there to be some Montreal inside Google.

Conferences

Somewhere in this period he became one of the people who ran the field’s institutions, and then became the person trying to change them.

He organized ICLR for several years and sat on its board, sat on ICML’s board, and eventually served as senior program chair of NeurIPS, the largest and most prestigious conference in machine learning, by then an event with thousands of submissions and an army of reviewers who had to be recruited, evaluated, and replaced. With the conference co-program chairs, he built the reviewer pool by pulling in everyone who had been an author on a recent NeurIPS paper, scored the reviews people wrote so the worst ones would not be invited back, and capped each reviewer at six papers.

After years of these conferences, Hugo had concluded that a conference is asked to do two jobs at once and does both badly. The first is to decide whether a paper’s claims are actually supported by its evidence, which is a question with an answer. The second is to decide which of the papers that clear that bar are exciting enough to feature, which is a matter of taste, and entirely legitimate for an event that exists partly to put on a show. Stapling the two together produces most of the randomness in accept and reject decisions, and it produces the particular misery of an author who cannot work out why a reviewer is unenthusiastic about work nobody has claimed is wrong.

So he launched a venue that would do only the first job. Transactions on Machine Learning Research launched in 2022, with Hugo and Kyunghyun Cho and Raia Hadsell among its editors in chief and Fabian Pedregosa as its founding managing editor, reviewing in the open against a single criterion, which is whether the claims match the evidence. Excitement is handled separately, through certifications attached to papers that have already been accepted, and the ambition is that conferences will eventually award those certifications themselves, so that a paper can be checked once and then featured by whichever communities find it interesting. The first venues to try it were small, AutoML and CoLLAs. And now, ICLR, ICML, and NeurIPS allow for some TMLR papers to be presented at these conferences, and Hugo feels quite proud of having made that possible. Submissions are accepted at any time, which dissolves the deadline cycle that has every researcher in the field cramming toward the same four dates a year and running out of compute in the same week. It reached roughly a thousand submissions a year, and a year later, Hugo said it no longer felt like an experiment.

Mila

What ended his time at Google was publishing.

He had joined on the understanding that he would be free to publish, and for years that was uncontroversial, because Google was widely regarded as the place to do AI and had concluded it gained more than it lost by putting its research into the world. Then OpenAI released ChatGPT and demonstrated that somebody else might reach very powerful systems first, and Google grew steadily more careful about what it was willing to say in public. Hugo does not dispute the company’s right to make that call. It is a company, and it is up to a company to decide. It was simply no longer the job he had taken. He went part time so he could spend his own hours at Mila, not as a Google person but as a passion project on his own dime.

Two other things were happening at once. The relationship between Canada and the United States had soured, and he found it less appealing to be contributing to an American company, though he is careful to say Google always felt more international than American and that this alone would not have moved him. And Yoshua Bengio had decided to hand off the scientific direction of Mila so he could put his energy into LawZero, his AI safety project.

The professors at Mila approached him, and Hugo said yes to join Mila as the successor of Bengio in the role of the Scientific Director.

The role is the one he has been doing for a decade, which is gardening, except the garden is now an institute of more than 1,500 researchers, including students, with a community of associate professors who work at the intersection of AI and other disciplines and whom he treats as his own continuing education. When people ask whether it is intimidating to take Yoshua’s chair, he says it is an honour, that the next chapter is a different one, and that what he brings to it is industrial.

What he took from Google was not a technique but a posture, the ambition, the insistence that a discovery is supposed to become something real, and the observation that Canada has been a world leader in AI research and considerably worse at putting any of it to use. He wants Mila’s students to graduate feeling capable of starting companies. He wants to narrow the distance between what the research can do and what anyone actually uses, and he wants it narrowed in a way that is recognizably Quebec’s rather than Silicon Valley’s.

Vigilant

At home there are four daughters, and what they hear at school is that AI is something they are not supposed to touch. He thinks that is a conversation Quebec is going to lose. The rule should be about cheating, which is a rule that already exists and always has, and which does not change because the tool changed. So he keeps it unforbidden in his own house, telling his children that a question they are wondering about is one they could put to Gemini, pulling out his phone in social situations to see whether a machine provides a useful suggestion, encouraging them to use it when they are stuck on a concept and want to understand it rather than when they want to reach the answer quickly.

The word he uses when people ask whether he is worried is “vigilant,” and he says it is deliberate and that not enough people use it. The opportunities are real, the risks are real, and which of them you end up with depends on how the technology is adopted, so the right posture is constant attention. He thinks the field has done better here than social media did, in that the scrutiny arrived early rather than a decade late, and he considers that healthy.

The idea he keeps returning to lately comes from François Jacob, who divided scientific work into day science and night science. Day science is coding up the idea, running the experiments, and analyzing the results. Night science is deciding which problems are worth years of a life, forming hypotheses, choosing where to point everything.

Hugo has watched coding agents tear through day science at a speed that has unsettled his own students, who are asking, with some justification, whether what they are doing is research at all. He is comfortable with that part being automated, because the middle of a project was never the fun part. The joy is at the beginning, when the idea arrives, and at the end, when it turns out to have been a good one.

Night science is different, and his objection to automating it has nothing to do with whether a machine could. It is about legitimacy, about who is entitled to decide where a society spends its time, money and attention, and about wanting a person accountable for choices that large.

He thinks the field does not do enough of it. He includes himself, saying he has mostly done night science when the job demanded it, that he could be more ambitious about it, and that it is the part of the work he finds most exciting anyway.

He talks now about wanting Mila to be where the next real breakthrough happens, and he is honest that the past decade has mostly been the scaling of ideas worked out while he was a graduate student. He cannot make that breakthrough arrive alone.

What he can do is keep the room open, the same way it was open for a student who was turned away by Bengio for not having enough mathematics and came back the next year.

Today we’re announcing our investment in Galaxia Space Systems, a Halifax-built company making it possible for any organization to run a space mission, without ever owning a satellite. Amiral Ventures is proud to lead Galaxia’s seed round alongside Invest Nova Scotia.

Space just crossed its economic threshold

For most of its history, the space industry was defined by a single number: the cost of putting one kilogram into orbit. In the Space Shuttle era, that figure sat in the tens of thousands of dollars — high enough that only governments and telecom giants could afford to operate above the atmosphere. Reusable rockets broke that constraint. Over the past decade, the cost of reaching low Earth orbit has fallen by roughly 90%, to a few thousand dollars per kilogram, and next-generation launch vehicles are targeting another order-of-magnitude reduction.

When the price of a scarce input collapses, entire industries get rebuilt around abundance. That is what is happening above us right now. Research from the World Economic Forum and McKinsey projects that the global space economy will nearly triple, from $630 billion in 2023 to $1.8 trillion by 2035, roughly twice the pace of global GDP growth. Notably, most of that growth won’t come from rockets or exploration. It will come from satellite-enabled services: Earth observation, communications, positioning, and the orbital data layer that industries from agriculture to logistics to defence increasingly depend on.

But cheaper rockets only solved half the problem: getting things into orbit is now easy; getting intelligence back out of it is not.

The bottleneck has moved from the rocket to the data

Today’s satellites are, for the most part, cameras with radios. They collect a massive amount of raw data in orbit, but sometimes require several passes over a ground station to downlink everything, and only then does analysis begin. The result is an intelligence cycle measured in hours or days. This is unworkable for defence and security applications that need answers in seconds, and increasingly limiting for commercial users monitoring ships, pipelines, wildfires, and infrastructure in real time.

There is a second constraint hiding in plain sight: traditional satellites are frozen at launch. Their software cannot be meaningfully updated, their mission cannot adapt, and every new use case requires designing, building, and launching a new spacecraft. Imagine buying a server that could only ever run the application it shipped with. That is how most of the orbital layer works today.

The lesson of the last two decades of computing is that infrastructure wins when it becomes programmable. Space is next.

Why Galaxia matters

Galaxia is doing for space what the cloud did for computing: owning the physical infrastructure and providing the software environment that lets anyone build on top of it.

The company designs and manufactures roughly 80% of its satellites in-house in Halifax, including its onboard computers, power systems, and software-defined radio and wraps that hardware in MissionOne, a platform that delivers space access as a service. Clients can commission a full mission lifecycle, fly their hardware as a hosted payload on Galaxia’s buses, or, most importantly, use Virtual Satellite-as-a-Service (VSaS): deploying containerized AI applications to virtualized satellite instances already in orbit, and paying only for what they use.

Underneath VSaaS sits Mobius, Galaxia’s software-defined satellite platform. Mobius turns a satellite into a reprogrammable orbital compute node: applications can be deployed, updated, and retired in space the way they are in the cloud, and AI inference runs directly onboard, compressing the intelligence cycle from hours to seconds by analyzing data where it is captured, before it ever touches the ground.

None of this is theoretical. Galaxia has launched two satellites, is generating real revenue from the Canadian government (notably with the Department of National Defence), and will shortly launch four more spacecrafts. Furthermore, the commercial opportunity is what makes VSaaS compelling: any organization with an algorithm and a problem visible from orbit becomes a potential customer. An energy company could deploy a leak-detection model to a satellite already overhead and get pipeline alerts in near real time, no spacecraft to build, no mission to fund, just an application pushed to orbit the way software is pushed to the cloud today. Few seed-stage companies anywhere can point to two anchor government customers, hardware with flight heritage, and a commercial wedge that is scalable.

Sovereignty is structural

There is a second dimension to Galaxia’s position that we believe is under appreciated: it is one of the very few full-stack satellite platforms in the world that is entirely free of ITAR restrictions, the U.S. export-control rules that legally bar American providers from serving large portions of the global market.

That constraint matters more every year. Allied militaries, neutral nations, and commercial operators across Latin America, the Middle East, and Africa have genuine Earth-observation and intelligence requirements and are structurally cut off from most high-capability U.S. providers. NATO members are actively seeking sovereign or allied space capabilities. Canada itself faces a significant surveillance and communications gap across its Arctic and maritime approaches. For all of these buyers, an ITAR-free, Canadian-built platform is not a nice-to-have; it is often the only credible option. No amount of competitor capital can replicate that advantage.

The policy environment at home reinforces it. Canada’s 2025 budget committed $81.8 billion to defence over five years, anchored by a “Buy Canadian” industrial strategy, and BDC has launched a $4 billion platform dedicated to defence and national-security companies. A sovereign, dual-use space champion built in Atlantic Canada sits squarely in the path of that capital.

The team

Galaxia is led by founder and CEO Arad Gharagozli, who built and launched LORIS, Atlantic Canada’s first satellite, while founding the Dalhousie Space Systems Lab, and has spent a decade in mission-critical space computing. The core engineering team has been with him since that first launch, and the company has earned repeated institutional trust from both the CSA and DND. Through our diligence, the references were unusually consistent: the technical authorities who have worked most closely with Galaxia described a team that delivers ahead of schedule and beyond scope, and led by a founder who learns at a pace that stood out even to the people who trained him.

Space is an unforgiving domain, and hard-won operational experience, including the lessons of real in-orbit operations, is precisely what separates the high performing teams. Galaxia has that experience, and it shows in the discipline of its next-generation missions.

Why we invested

At Amiral Ventures, we back founders building the infrastructure of a more productive and more resilient economy. Few investments express that thesis as literally as this one. For a country like Canada, resilience increasingly means sovereign capability: the ability to see, sense, and decide using infrastructure it owns, in a world where access to allied systems can no longer be taken for granted. Galaxia is building that capability, and building it as a commercial platform rather than a cost centre.

We believe the orbital layer is at the same inflection point the internet reached when computing became a utility: launch is cheap, demand for real-time intelligence is compounding, and the winners will be the platforms that make space programmable. Galaxia combines the full hardware stack, the software environment, the government validation, and the structural market access to become that platform for Canada and the ITAR-constrained world.

As a Canadian fund focused on backing strategic technology companies built here, we are committed to supporting Galaxia with the network, follow-on capital, and enterprise and defence connectivity needed to compete at global scale.

Congratulations to Arad and the entire Galaxia team. We are excited to be on board as you bring the cloud to orbit.

In January 1993, an episode of The Simpsons aired in which a smooth-talking stranger in a straw hat, Lyle Lanley, arrived in Springfield with a song. The town had just come into a windfall of $3 million from a fine levied against the local nuclear plant, and the citizens were debating, at a town hall meeting, what to do with the money. Marge Simpson stood up to suggest they fix the cracked and crumbling Main Street. The crowd was unconvinced. Then Lanley took the stage and launched into a Broadway-style musical about how Springfield needed a monorail. Lanley promised modernity, glamour, and a place for the town in the ranks of Brockway, Ogdenville, and North Haverbrook as one of the great monorail towns of the American republic. The crowd, hypnotized, voted to give Lanley the money. The monorail was built but turned out to be a deathtrap, and Lanley fled to Tahiti with a suitcase full of cash. The episode became one of the most quoted in the entire run of the series, and in certain corners of the policy world it has lived on as a stand-in for any too-good-to-be-true infrastructure pitch sold to a gullible government by a charismatic outsider.

John did not understand, in 1993, why this episode mattered. But it later became an episode that reflected his purpose in life.

Childhood

John Ruffolo’s father came from southern Italy. He had grown up on a watermelon farm and was destined, if he stayed, to spend his life on that farm. His mother came from central Italy. They both left as teenagers, in the great post-war migration that brought hundreds of thousands of Italians to the cities of southern Ontario in the 1950s and 1960s, and they met not in Italy but in Toronto, in the streets of the original Little Italy. They married, and moved into the working-class suburbs absorbing the Italian community as it grew, ending up in the corridor between Jane and Sheppard and Jane and Finch. They had three children. The eldest was a brother who would eventually work for the Toronto Transit Commission; the middle was a sister who would eventually work in retail; and John, the youngest, would eventually become the outlier.

The neighbourhood was an outpost of southern and central Italy transplanted onto the Canadian Shield. The gardens in every backyard grew tomatoes, zucchini, peppers and basil. The fences between properties had small gates cut into them so that neighbours could move freely from one yard to another and trade what they had grown, one family taking the tomatoes and another the peppers. The men worked construction or in the trades. No Canadian employer in those years would hire an Italian for anything else. Many of the women never learned English, because the world they needed to survive inside a few square blocks. When John was small, he believed he lived in Italy. It took him until he was 5 to figure out he did not.

The family was of modest means. Money came in on Friday and was gone by Wednesday. They ate what they grew. They drank wine they had crushed themselves in the basement, the kids pulled in to help with the crushing, their fingers stained purple for a week afterwards in a colour that would not wash off no matter how hard it was scrubbed. In September, when the tomatoes came in, every family on the block mobilized for the great annual canning of the sauce, and John would arrive at school the next morning with tomato seeds clinging to his hair, sleeves, and collar, and the seeds would still be there a week later, dropping off one by one. He found the seeds humiliating, but the wine fingers were worse. He spent his entire childhood trying to get out from under both.

He had a vast extended family of cousins, aunts and uncles arrayed around him on every side, and within that extended family, which he describes as enormous, he was the first person ever to attend post-secondary school. The rest of the family had gone into the trades. They were tile setters, cement workers, painters, and small entrepreneurs running floor businesses or grocery stores. They were entrepreneurs not because they had chosen entrepreneurship but because the doors to salaried employment had been closed to them. From this, his parents drew the opposite lesson: do not become an entrepreneur. Do not work with your hands, do not run your own business unless you are forced to, become a professional and wear a suit and earn a salary, become a lawyer or an accountant or an engineer or a doctor. The professions, in their reading of the Canadian economy, were the only category of work that came with the credentialed legitimacy that no amount of prejudice could override. John absorbed this and accepted it.

By the time he was 6 he was reading Martin’s Criminal Code, the standard Canadian legal reference, in his bedroom, having decided he was going to be a lawyer and that the law was therefore something he should probably understand.

His father, when John would later ask him as an adult why he had left Italy if he loved Italy so much, would yell at him. “You think you would be doing what you’re doing now, enjoying your life, if I stayed there? You would have been working with me on the farm!” John would concede that this was a reasonable answer.

His earliest hero, oddly, was Pope John Paul II, not because the family was particularly devout, although they were Roman Catholic, but because the man on television was the first Pope in living memory to leave the Vatican and walk among the people, and the small boy watching from his living room recognized in this figure something he could not yet name. He would later be drawn to the Dalai Lama for the same reason. To Nelson Mandela, who walked out of 27 years in prison and refused to grow visibly angry. To Muhammad Ali, who had no formal education, became one of the great communicators of the 20th century anyway, absorbed beating after beating in the ring and beating after beating in his political life, and refused to blink. The pattern running through every figure John found himself idolizing was the pattern of the maverick, the non-conformist, the person who had been told the thing they were trying to do was impossible and went ahead and did it anyway.

The Bank at Jane and Finch

When he was 16, John got a job as a teller at the Bank of Montreal branch on the corner of Jane and Sheppard. He took it for the money and the chance to wear a tie. Within a year he had been promoted to manager, on Thursdays, Fridays and Saturdays, of the Bank of Montreal branch at Jane and Finch.

The branch at Jane and Finch was famous in Canadian banking for two superlatives, both unwanted. It had the highest number of daily transactions of any branch in the entire Bank of Montreal system, because it served a community of working-class immigrants and pensioners who came in person every day to deposit their cheques, pay their bills and send money home. It had, by a wide margin, the lowest average dollar value per transaction of any branch in the entire system, for the same reason. The neighbourhood was rough in a way that Toronto polite society did not like to acknowledge. Anyone with half a brain and any career to protect had said no when offered the job of running it, and the bank had eventually worked its way down the list to a 17 year old high school student in a borrowed suit. He said yes immediately. He did not understand he was supposed to say no. “Of course I’m going to do that,” he said years later. “I’m the naive guy.”

What he learned in that branch, in the long Saturday shifts when he was the senior person in the room and entire families would come in to argue with him in three different languages about a hold on a cheque or a bounced money order, was the entire architecture of finance from the inside out. He learned what cash flow meant when it belonged to people with no margin for error. He learned what a default looked like in person, sitting across a desk from someone whose face you knew from the church on Sunday. He learned that finance was not, fundamentally, about numbers. The numbers were trace evidence left behind by human decisions made under pressure. The way to be good at finance was to understand the people first and the numbers second. By the time he finished Grade 12, the lawyer ambition was gone, replaced by something he did not yet have a name for but that pointed in the direction of Bay Street.

He went to the Schulich School of Business at York University, which in those years was the natural destination for ambitious working-class kids from the Toronto suburbs who could not afford to look further afield. In his third year summer he was wrestling between two offers. The first was from Bank of Montreal, in what was then called corporate and government banking, the predecessor of investment banking. The second was from the Toronto office of Arthur Andersen, the great accounting and consulting firm. He could not pick. So, with the unreasonable energy of 20-year-olds, he simply did both jobs, 80 hours a week for four months, and let the firms compete for him in real time. The decision came down to a single conversation. The senior banker told him that if he stuck with the program and worked hard for twenty years he might one day occupy a position similar to the vice president’s. Twenty years sounded like the entire span of his life again. The Andersen partner, asked the same question, said the firm would let him move as fast as he wanted. There was no ceiling and no schedule, only the speed at which he himself was willing to run.

“Oh,” John thought. “I will take that.”

Arthur Andersen

What awaited him on Bay Street was a culture shock Schulich had not prepared him for. Arthur Andersen in the late 1980s was the most prestigious of the Big Five accounting firms, and its Toronto office was populated by graduates from premier private schools of the Canadian establishment, whose families had been on Bay Street for generations and knew each other already from sailing camps in Muskoka. He had never met any of these people. He had a chip on his shoulder, an Italian surname and a hairy chest. Within his first weeks at the firm a colleague walked into the men’s locker room while he was changing for a sports league and said that he must have big feet too, using a slur for Italians we will not reproduce here. In any other context John would have punched the man in the face. Instead he filed it away. He understood the kind of room he had walked into.

He moved from audit, where he had started, into tax, because tax was where the deals happened. He had developed, in his short time on Bay Street, a deep affection for the choreography of mergers and acquisitions and IPOs. By 1992, at 25, he had not yet made partner. The country was in the depths of a recession. He was bored, trained as an international tax expert, but with no transactions to do international tax on. The clients he was being handed were giving him their tax problems, and the part of the work he loved was not the tax. It was sitting across from the founders and asking them to explain how their businesses actually worked. He could do that for hours. He noticed, with some surprise, that almost no other tax person on Bay Street did it at all.

He went to the partners and asked permission to build a new practice within the firm dedicated to serving Canadian companies in a particular industry. He had thought about which industry. The banks, manufacturers, and the resource companies were obvious candidates, but the people running them were a generation older than he was, and his odds of getting a meeting with a 55-year-old CEO as a 25-year-old manager were close to zero. The technology industry, by contrast, had founders close to his own age. They would not, he reasoned, treat him like a piece of crap simply because he was 25. This turned out to be exactly correct, but the bet took years to pay off. The senior partners, who did not particularly care about technology one way or the other and who had no expectation that someone John’s age could realistically build a practice from scratch, gave him their blessing the way they gave blessings to children who want to dig for gold in the backyard. “Knock yourself out,” they said.

He had been handed two technology clients early on, Oracle and Microsoft, and from those two relationships he started knocking on the doors of every founder in Toronto who would meet with him. Total revenues for the technology practice, Year one: $0. Year two: $0. Year three: $0. He was, all the while, doing his full tax workload on top, billing the same 60 hours a week he had been billing before, because if the technology practice continued to bill nothing his only insurance against being fired was the tax revenue he kept generating on the side. He was running two careers in parallel, and he was tired.

The break came in late 1994, from an unexpected direction. A team of engineers at the National Center for Supercomputing Applications in Illinois, including a young man named Marc Andreessen, released the Mosaic browser, which they spun out into a startup called Netscape. The browser made the consumer internet visible, for the first time, to ordinary people. Within months every major company in North America was scrambling to figure out what its internet strategy was going to be. Within a year the entire landscape of business had been reorganized around a category of company that, six months earlier, had not existed. John, who had spent three years building a practice nobody at his firm believed in, was suddenly one of the only senior people at the only major accounting firm in Toronto with deep relationships across the technology industry, just as the technology industry was about to become the industry. The fees, for years a flat zero, started arriving in the millions. He was on fire and his career took off.

He remembers a partner walking up to him in this period and said, in a tone of weary congratulation, that “it must be pretty lucky how the technology thing had worked out.” He looked at the partner and thought, “Are you kidding me?! You threw me to the dogs for three years and now I’m making more money than you are and you call it luck?” Obviously, he did not say it out loud.

Partnership Shmartnership

By the autumn of 2001 he had been promoted into the global partnership of Arthur Andersen, the small inner sanctum of the firm whose members held an interest in the worldwide partnership in Switzerland, not merely in their local national practices. He was the youngest partner at the time of his elevation. There was a small administrative matter to attend to before he could take up his new seat. Becoming a global partner at Andersen required a capital contribution. Specifically, it required a personal cheque for $1 million.

The son of an Italian house painter from Jane and Finch did not have anything close to a million dollars in 2001 currency. The firm had a standing arrangement with a group of friendly lenders who would loan a new partner the buy-in amount against the security of the partnership interest itself, and so John borrowed the full seven-figures, signed it over, and walked back to his desk a global partner. Three weeks later, the news broke about a small accounting irregularity at one of Arthur Andersen’s largest clients, an energy trading company in Houston called Enron.

What followed has been chronicled in books, films and graduate school case studies. What is harder to convey is what it felt like to be a brand new global partner at the firm whose name was on top of the disaster. The Enron lawsuit was filed in the United States, where the audit work had been done, but because John had just bought into the global partnership his interest was directly exposed to the U.S. firm’s liability. There were no limited liability protections in place at the time for partnerships of this kind. If the U.S. firm was found liable for the multi-billion-dollar damages being claimed, the global partners around the world were on the hook personally, with no cap. John faced the loss of every dollar he had ever earned, the forced sale of his house, and personal bankruptcy. Bankruptcy carried with it, under the rules of the Chartered Accountants of Ontario, the automatic loss of his CPA license. Inside of nine weeks he had borrowed a million dollars he did not have, lost it on paper, and discovered he was on track to lose his career as well.

His marriage was failing at the same time. The dot-com market, which had been the engine of his rise inside the firm, had begun its long collapse the previous year and was now, with the World Trade Center attacks layered on top, in something close to free fall. The global partners voluntarily cut their incomes by two-thirds overnight. On a personal balance sheet that was already underwater this was an act not of generosity but of grim necessity. He was finished.

Instead of breaking, he moved out of his marital house and rented a room on the top floor of a home elsewhere in his old neighbourhood, kept his bicycle, went to his favourite coffee shops and bars, and had pasta at his mother’s on Sundays. After a couple of months of this he sat down with himself, in one of those coffee shops, and ran the same logic he had run at Jane and Finch. What was true was that he was alive, that he was healthy, that he could still ride a bicycle, drink a cup of coffee, and be loved by his family. The amount of income he was earning, even on one third of his previous salary, was extraordinary by the standards of the world he had grown up in. He was experiencing a first-world problem. His parents, who had crossed an ocean with nothing in their pockets, would not have recognized it as a problem at all. “If this is the worst it becomes,” he said to himself, “this is no problem.” Hope, he later said, touches the same door as despair. He chose the door of hope.

The postscript nobody saw coming was that none of the threats came true. Arthur Andersen, in the global wreckage of Enron, was rolled into Deloitte in Canada. The merger was structured so that the global partners’ liability was fully contained and never crystallized. The capital contributions were fully refunded. He got his million dollars back, kept his CPA license, and avoided bankruptcy entirely. The merger with Deloitte, which had appeared in the moment to be an act of corporate humiliation, turned out to be the slingshot that launched the second phase of his career. He arrived at Deloitte in 2002 with the half-feral instincts of an Andersen survivor and the resources of the largest accounting firm in Canada at his disposal. He had been broken almost to the point of failure and what came out the other side was a man with the confidence of someone who had already lost everything once and watched the world keep turning anyway.

Pension Funds

The Deloitte years ran from 2002 to 2010. He took over the firm’s global leadership for technology, media and telecommunications, which gave him a perch from which to observe the entire arc of the industry on every continent. He served as the personal advisor to the CEOs of Rogers, BlackBerry, and Open Text. He sat on the board of Deloitte itself.

By around 2005 he had begun to develop a worry that no amount of personal success could quiet. The dot-com crash had not really receded. It had simply settled in. Venture capital in Canada, which had never been particularly robust, had hollowed out. The domestic venture firms of the 1990s had folded or were running out of money on their final funds. The pension funds, which in any other country would have been the natural source of replacement capital, had publicly declared themselves uninterested in the asset class and were diverting their allocations into infrastructure, real estate, and private equity buyouts of mature businesses. The kids coming out of the universities were leaving for Silicon Valley at a rate that was not sustainable for any country that hoped to compete in the digital economy. Canada was on track to lose an entire industrial generation.

He decided to put his shoulder against a problem that was not strictly his problem. He spent the next three years working on public policy, advising the federal and provincial governments on what could be done to revive the venture capital ecosystem. He ran the government relations committee at the Canadian Venture Capital Association. He wrote the first draft of what would eventually become the federal Venture Capital Action Plan. He gave speeches, wrote papers, and began to publicly attack the pension funds. He attacked them in private meetings with their CEOs and CIOs, and when those meetings produced nothing he attacked them in the financial press, in industry conferences, in op-eds. He proposed a 1% pension tax, a mandatory levy that would force the large funds to contribute to a domestic innovation pool whether they wanted to or not. The pension fund executives, who had not previously considered themselves the sort of people who got attacked publicly by Bay Street accountants, did not enjoy this. They responded badly, and he kept attacking.

In late 2010 the telephone rang in his office at Deloitte. It was Michael Nobrega, the chief executive of OMERS, the Ontario Municipal Employees Retirement System, one of the five largest pension funds in Canada. John knew Nobrega from a long way back. Nobrega had been a senior partner at Arthur Andersen during the summer when John had been a tax intern there, and the two had stayed in loose touch in the decades since. The call, John assumed, was going to be a polite invitation to stop attacking pension funds in print.

It was not. Nobrega told him OMERS was thinking about getting into innovation. He asked John whether he would consider serving as an advisor to help structure the program. “Oh my god,” John said. “That’s wonderful.” Three meetings later the request had escalated. Nobrega did not want him as an advisor. Nobrega wanted him to come over and build a venture fund. “We don’t know what to do,” Nobrega said. “You have been poking us. You think you’re so smart. Why don’t you do it?” John, who was a partner on Deloitte’s national board and was being seriously considered as a future leader of the practice, found himself trying to talk Nobrega out of the offer. He laid down conditions, asking for things he assumed OMERS could not deliver, but Nobrega kept saying yes. He had run out of obstacles. With one last bit of risk mitigation he asked Deloitte to insert a clause in his resignation papers stating that if he chose to return to the firm at any point in the future he would be deemed never to have left. Deloitte agreed.

He arrived at OMERS in January of 2011. On the 1st of December, 2010, less than a month before his start date, an opinion piece appeared in the Globe and Mail. It was written by a Canadian venture capitalist in the Valley, a “friend” of John, and it argued in not particularly veiled terms that OMERS had made a catastrophic mistake by hiring John. It pointed out he had never invested capital in his life and questioned whether an accountant from Bay Street could possibly build a venture capital firm. The piece was particularly galling because John had spent the previous five years at Deloitte advising essentially every venture capital firm in Canada and several in Silicon Valley, sending them deal flow, doing diligence work for them, and feeding the author of the article in particular more business than he had probably gotten from any single source. He found out later that the author had wanted the OMERS job himself. The Globe piece was the public expression of professional resentment. He read it, then received a call from Nobrega to which he said, “John, I’ve got your back.” The chip on the shoulder, quiet for some years, came roaring back. He was going to prove every one of them wrong, in public, with money that was not his.

OMERS

The first few years at OMERS Ventures were a period of intense and largely invisible work. John and his small team focused initially on the smallest stage of the market, on companies between $0 to $10 million of revenue, because that was the segment where the lack of domestic capital was most acute. The first investment they made was in a Toronto-based company called Wave Accounting, which had built a free online accounting platform for very small businesses. Over the following decade Wave would return roughly 100x the original capital. Anyone with the guts to write a cheque in 2011 and the patience to hold the position was going to make money. What was difficult was the guts and the patience.

By 2012, with the first wave of small investments out the door, he started seeing something else move through the market. A handful of his early portfolio companies were growing far faster than he had expected, blowing through the $10 million revenue mark inside of two years and arriving at a stage where they needed a much larger cheque to keep scaling. Nobody in Canada in those years could write that cheque. The entire growth-stage segment was a desert. American firms could fly in from Boston and Silicon Valley and write the cheque, but they would extract, as the price of the capital, a significant migration of the company’s center of gravity to the United States. This would be the inevitable consequence that the eventual exit and wealth creation would happen below the border. He decided to push OMERS Ventures up the stage curve.

In the summer of 2013 there was a coffee chat that turned into the defining transaction of his career.

The man on the other side of the table was a soft-spoken German immigrant named Tobias Lutke. Tobi had moved to Ottawa in pursuit of a Canadian woman he had fallen in love with. In Ottawa he had co-founded a small online snowboard shop called Snowdevil. While trying to build the website for the shop he had become so frustrated by the available e-commerce software that he had ended up writing his own platform from scratch, in a programming language called Ruby. His friends, who came in to buy snowboards, started asking where he had gotten the website software, because they wanted to use it for their own little online businesses. Tobi pivoted away from snowboards entirely and rebuilt the company around the platform itself, which he called Shopify. By 2013 the company was growing fast and looking to raise the largest financing round in Canadian technology history, $100M.

John walked into the meeting expecting to spend half an hour on the company’s financials, growth model and competitive position. What he got instead was a 2.5 hour conversation in which the topic of Shopify, in the narrow business sense, almost never came up. Lutke talked about artisans. He talked about small businesses, about the fundamental injustice of a world in which a cabinetmaker in a small town or a leather worker in a back-alley studio could not, simply because they did not have the technical skills to build their own e-commerce platform, sell their wares to the world. He talked about why he had moved to Canada, why he loved the country. The whole conversation came out of him with the undiluted conviction of a man who had identified one particular problem and decided it was the central project of his life.

By the end of the meeting John had decided, more or less without having looked at the company’s books, that he was going to write the cheque. He went back to his team and told them he had no idea whether the software actually worked but they had to back this founder. They put the deal together. OMERS led half of the $100M. Shopify went public 20 months later, at a valuation north of a billion dollars. It has since become the largest publicly-traded company in the history of Canada and the most important wealth creation event in a generation of Canadian technology.

The single most important thing about the Shopify investment was not the software. The software was good and the market was big, but neither of those things was what mattered; what mattered was the founder. “Don’t chase the money, don’t chase the glory, don’t chase the fame,” he says. “Chase the problem. The problem will get you around every obstacle.” He came to think of his work not as financial work but as pattern recognition work. The technology underneath any given company was going to change, sometimes within months, with the market shifting alongside it and the product shifting again on top of that. What did not change was the founder’s relationship to the problem. Whether they would still be obsessed with that problem ten years later, after they had become rich and famous and had every excuse to walk away, was the only question that mattered.

Around Shopify, the OMERS Ventures portfolio filled in. There was Hopper, the Montreal-based travel booking application that used machine learning to predict flight prices; Rover, the Seattle-based marketplace for dog walking and pet sitting; Wattpad, the Toronto-based storytelling platform that grew into one of the largest reading communities in the world; DuckDuckGo, the privacy-first search engine built as a deliberate alternative to Google; League, the Toronto-based digital health benefits platform; TouchBistro, the Toronto-based point-of-sale software for restaurants; and Xanadu Quantum Technologies, the Toronto-based pioneer in photonic quantum computing. By the time he finished his run at the firm, OMERS Ventures had deployed more than $500M across more than 40 companies. In parallel, almost by accident, he had opened a co-working space at 111 Richmond Street West that became the clubhouse of the new Toronto technology scene. Founders, operators and investors collided in the hallways. The room ran on what he calls coopetition. When founders raised money or had an exit, the rest of the community cheered. They were not pretending when they cheered, and the culture he had built around that fact was the rarest thing in Canadian business.

He has, over the years, been candid about the failures alongside the successes. The pattern in the ones that did not work, looking back, is consistent. Either he had misjudged the founder’s resilience, or he had failed to be honest about a follow-on. The first failure mode was about purpose. The founders who fold under stress are not the ones who fold loudly. They are the ones who, when things get difficult, quietly say “I’ll try something easier” and pivot away from the original problem. The founders who survive are the ones who refuse to pivot away from the problem because the problem is the reason they got out of bed in the first place. The second failure mode was about ego. Venture investors, in his experience, do a reasonably good job on their first cheque into a company and a terrible job on the follow-on. The follow-on is where the ego protects itself. Investors convince themselves that doubling down is loyalty when it is in fact denial. He instituted, both at OMERS and later at Maverix, a strict policy that every follow-on had to be evaluated as if it were a brand-new investment, with no credit given to the prior conviction.

The framework underneath all of this, the larger argument that would eventually subsume the venture capital work and reorganize the rest of his life, was still missing. It arrived in 2014, in a book.

Peter Zeihan

The book was The Accidental Superpower, by an American geopolitical analyst named Peter Zeihan. Zeihan had spent the early part of his career at Stratfor, the private intelligence firm, before going independent. His method, refined across what would become a series of four books, was to start from two variables that almost no other geopolitical thinker treated as primary: geography and demography. Everything else was downstream of those two.

The argument of the book was that the United States was not a great power because of its political system, military, or culture, but because of its geography. The country sat on the largest contiguous stretch of arable temperate farmland on the planet, irrigated by the most extensive and navigable internal river system in the world. It had two ocean moats protecting it from the great powers of Eurasia, friendly neighbours to the north and south, deepwater ports on both coasts, and abundant domestic energy. The post-1945 global order had not been an act of generous statesmanship by the Americans, but a strategic bribe. The U.S. Navy would guarantee the freedom of the seas for everybody, friend or foe. The American consumer market would be opened to allied exports on extraordinarily generous terms. The U.S. dollar would become the world’s reserve currency. In exchange, the allies would line up against the Soviet Union and accept American leadership of the Western alliance. The German export miracle, the Japanese industrial revival, the rise of the Asian tigers, the Chinese manufacturing boom, all of it had been made possible by the American naval umbrella and the American consumer market. None of it would survive the unwinding of the deal, which was already underway. The Soviet Union had collapsed in 1991. There was no longer a strategic reason to bribe the rest of the world. The shale revolution, which had begun in the late 2000s, made the United States energy-independent, which meant the Americans no longer needed Middle Eastern oil. American manufacturing had been hollowed out by the very globalization Bretton Woods had made possible. The political coalition that sustained the order had collapsed at home. What was coming was a great unwinding. Long supply chains across multiple oceans only worked if somebody policed the oceans for free. Just-in-time manufacturing across continents only worked if shipping was cheap and predictable. Without the American naval umbrella, all of it would fragment. Energy markets would regionalize. Trade routes would become contested. The countries that depended on imported food and imported energy, which was most of the developed world outside North America, would find themselves vulnerable in ways they had not been since 1945.

John read the book in 2014 and something clicked. He gives, as an example, the question of which Canadian technology has produced the greatest global disorder of the 21st century. The answer, almost nobody guesses, is fracking. Fracking was developed in Canada, but was not patented or systematically commercialized. The Americans took the technology, built an entire industry around it, and used it to convert themselves from a net importer of Middle Eastern oil into a net global energy exporter, which gave them the ability to tell the rest of the world to fend for itself. One Canadian technology, never properly captured, reordered the global balance of power.

The Zeihan framework gave John a vocabulary that connected the things he had been working on as discrete problems for a decade. Capital, talent, trade rules, data, intellectual property, energy, food, and defense were not separate problems but the same problem viewed from different angles, all expressions of a single underlying question, whether a country controlled the conditions of its own economic and physical existence or had outsourced those conditions to somebody else. The word for the answer to that question was “sovereignty”. Zeihan, alongside Ian Bremmer at the Eurasia Group, whose Monday morning newsletter John’s entire team waits for and reads in full, has become 1 of 2 intellectual architects of the way the firm sees the world.

The Council

In October 2015, John and Jim Balsillie founded the Council of Canadian Innovators.

Jim, the former co-CEO of BlackBerry, had spent the years since the collapse of his company’s competitive position learning how American technology firms actually compete internationally. The lesson he came away with, and that he delivered in a speech at one of John’s CEO roundtables in 2015 that John remembers as the most important speech he ever heard, was this. When a Canadian technology company competes against an American technology company, the Canadian company is not actually competing against the American company. The Canadian company is competing against the entire United States government. American technology firms, especially the largest ones, have spent decades building lobbying and government relations infrastructure that allows them to shape the rules of the game in their favour. Trade treaties, tax legislation, intellectual property frameworks, regulatory standards, procurement policies, and antitrust enforcement. All of it is influenced, in ways that are usually invisible, by the industries it affects. The Americans do this masterfully, but the Canadians do not do it at all. Worse, the Canadian elite has developed a kind of moral self-flattery around its refusal to play this game. Canadians like to tell themselves they are above lobbying, above industrial policy, above the grubby work of writing rules in their own interest. They prefer to think of themselves as Boy Scouts in a world of free and open commerce that does not exist. They were written by American industries and American lawyers in American interests, and Canadian companies trying to scale globally are running into walls they do not even know are there.

By the time Jim finished speaking, the room of CEOs was depressed. John pulled him aside and asked what they could possibly do about it. Jim’s answer was that no individual Canadian technology company could afford the kind of government relations infrastructure American companies took for granted. A serious lobby shop costs millions a year and requires expertise Canadian founders, busy running their companies, do not have. But if the resources of 50 or 100 Canadian technology CEOs were pooled together, then it could be afforded. A single organization could be built that handled the government relations function for all of them, focused only on issues where their interests aligned, and gave them, for the first time, real political weight in Ottawa.

That was the founding insight. The Council of Canadian Innovators, with Benjamin Bergen as its first president, was built to be the shared lobbying arm of Canadian technology scale-ups, paid for entirely by member companies and answerable only to them. The membership criteria were narrow and deliberate. To join, a member had to be Canadian-headquartered, in the technology sector broadly defined, and generating meaningful revenue, and the narrowness of those criteria was the point. Jim and John did not want a generic business association diluted by the priorities of the banks and the telcos. They wanted a focused organization representing the specific interests of the people building the next generation of Canadian technology companies, because those interests were systematically underrepresented in Ottawa.

The substantive policy agenda has clustered around three themes John summarizes as access to capital, access to talent, and access to markets and customers. On capital, the Council has spent a decade hammering on the same point John was making in 2010, that Canadian pension funds and Canadian banks systematically underinvest in Canadian innovation. On talent, the Council has been deeply involved in immigration policy, particularly in pushing for fast-tracked visas for highly skilled technology workers; the federal Global Talent Stream programme, one of the rare unambiguous successes of recent Canadian industrial policy, was shaped in significant part by Council advocacy. On markets, the Council has fought a long and largely losing battle against Canadian government procurement practices that systematically favour American technology vendors over Canadian ones. The federal government, despite years of nominal commitments to “buying Canadian,” continues to spend the vast majority of its technology budget with American hyperscalers and American software vendors. The Council has argued, with John as one of its most public voices, that this is a sovereignty issue as well as an industrial policy one. Canadian government data sitting on American cloud infrastructure is subject to the U.S. CLOUD Act, which gives American courts the right to subpoena it without notifying the Canadian government. The Council has been making this argument since before most Canadian politicians had heard of the CLOUD Act, and it kept making it through every data sovereignty fight of the last decade.

What makes the Council distinctive, and what John keeps emphasizing, is its operating posture. Most Canadian business associations approach Ottawa as adversaries, criticizing government from outside and demanding things. The Council, by design, embedded itself inside the policy process. Its staff includes former civil servants and former political staffers from multiple parties. Its members brief Ministers and Deputy Ministers regularly. When the federal government needs to understand a technical issue affecting the technology sector, the Council is often the first call. When Silicon Valley Bank collapsed in March 2023 and threatened the cash positions of dozens of Canadian technology companies overnight, the Council was already advising the federal Department of Finance and the deputy prime minister’s office before most other industry groups had figured out what was happening. John tells this story with measured pride. Being inside the room, having earned access through 10 years of consistent technically credible engagement, is more valuable than being outside the room complaining.

Two years after the Council’s founding, the test case arrived. It came dressed as Lyle Lanley’s song.

Sidewalk Labs

In 2017 Waterfront Toronto, a tri-government agency representing the federal, provincial, and municipal governments, announced that a subsidiary of Google called Sidewalk Labs had won a competition to develop a 12-acre site on Toronto’s eastern waterfront called Quayside. Sidewalk Labs was run by Dan Doctoroff, the former deputy mayor of New York under Bloomberg, and was sold to the public as a kind of urban innovation laboratory. The vision, as marketed, was the world’s first truly smart city neighbourhood, built from the internet up. Sensors everywhere, adaptive traffic signals, heated sidewalks, modular buildings, autonomous vehicles, pneumatic waste collection. A digital layer running underneath the physical one, capturing data on every movement and interaction in the neighbourhood and using that data to optimize services in real time. The pitch promised hundreds of millions of dollars of investment, thousands of jobs, and a showcase development that would put Toronto on the map alongside Singapore and Songdo.

When John first read the proposal, he thought of The Simpsons. He thought of Lyle Lanley, his song, the moving sidewalks Sidewalk Labs had literally promised, and how Springfield’s monorail was going to be the talk of the country.

Jim had been seeing it longer than John, arguing for years that Google’s actual business model was not advertising. Google’s actual business model was the extraction of economic rents from any digital surface it could control. Search, email, mobile operating systems, app stores, mapping. In every domain Google entered, it positioned itself as the toll booth between the user and the service. Sidewalk Labs, in Jim’s reading, was an attempt to extend this model from the digital world into the physical one. Cities were the next surface. Whoever owned the data layer of a neighbourhood would own the neighbourhood. They would extract a toll on every movement, transaction, and interaction. They would capture the value of the autonomous vehicle revolution by owning the data on which autonomous vehicles depended. They would relegate Canadian innovators to the role of app developers in someone else’s app store, treading water on whatever crumbs were left over after the toll had been collected. The smart city, properly understood, was a platform owned by Google, and Canada was about to give them a Canadian neighbourhood as a free testbed for technology Google would then deploy globally.

John listened and ran the math. Through Maverix, he invested in a Kitchener-based intelligent traffic systems company called Miovision, which does exactly the kind of work Sidewalk Labs was proposing to capture. He understood, from Miovision’s perspective, what was at stake. Whoever owned the data on traffic flows in a city owned the foundation layer for autonomous vehicles. Whoever owned the foundation layer for autonomous vehicles owned a multi-trillion-dollar future industry. Sidewalk Labs was not asking for permission to build a neighbourhood. They were asking for permission to build the data platform on which the entire future of urban mobility would run, and they were asking for it, in exchange for a few hundred jobs and a press release, on the cheap.

John did something almost nobody else thought to do. He bid against them, putting together a competing proposal for the Quayside site, and came in as the runner-up. He had not actually expected to win; he was bidding to verify, from inside the process, whether Jim was right. From inside the process, he confirmed that Jim was indeed right. The deal as structured was a transfer of sovereignty over data, taxation, and regulatory authority from a Canadian government to an American corporation, dressed up as an investment.

He and Jim went to war. They wrote op-eds, briefed members of Parliament, and appeared on every panel and every podcast that would have them. The deal as proposed, they argued, would give Sidewalk Labs the right to capture portions of property tax revenue to fund its own infrastructure, which they called, accurately, privatized taxation. It would give Sidewalk Labs intellectual property rights over systems developed using Canadian public data. And it would let Sidewalk Labs design its own governance structures, granting the company quasi-governmental powers over a piece of Canadian sovereign territory, with limited accountability to elected officials. Sidewalk Labs and its allies fought back. They lobbied OMERS, where John was still working at the time, to pressure him into shutting up. He did not shut up.

At one point during this period, Eric Schmidt, the executive chairman of Google, flew to Toronto and asked for a meeting at John’s office. John remembers thinking, “Who am I? I am a nobody. Why is Eric Schmidt flying to Toronto to meet me for an hour?” Schmidt arrived. The meeting lasted for 57 minutes, and the two men had what John describes as a genuinely fascinating conversation about geopolitics, technology, and the long arc of where the world was heading. Schmidt was a brilliant man. They liked each other, and John was fairly impressed. In the last 3 minutes, Schmidt took out a business card, scratched out the corporate phone number, wrote his personal mobile on the back, slid the card across the desk, and said, “If you have any issue with Sidewalk Labs, I want you to call me. I’ll resolve it for you.’ Then Schmidt got up and left, and John sat at his desk holding the card. The card was not a gift but an instruction to stop. He doubled down.

The deal slowly fell apart over 2019 as the public scrutiny intensified. Sidewalk had to keep retreating from its initial proposals, with the footprint shrinking, the IP claims softening, and the governance structure repeatedly revised. Critics that John and Jim had emboldened, including the open government advocate Bianca Wylie and the former Information and Privacy Commissioner of Ontario, Ann Cavoukian, kept the pressure on. By early 2020 the project was on life support, and in May of 2020 Sidewalk Labs announced it was pulling out of Toronto entirely. They blamed Covid-19, but almost nobody believed them. The public opposition had made the original deal undeliverable, and the revised version was no longer interesting enough to Google to be worth the fight. A year later Google wound down Sidewalk Labs as a standalone company entirely.

John asks, in the years since, a single question of anyone who still thinks Sidewalk Labs was a good idea Toronto should have embraced. “How many other cities in the world are currently doing what Sidewalk Labs was proposing here? How many other cities partnered with them after Toronto turned them down? The answer is none. Zero.” San Francisco refused. New York refused. Hudson Yards, the original American testbed, refused. Every other city in the world had looked at the proposal and said no. Toronto, alone among the great cities of the developed world, had looked at it and said yes. “We were the only idiots,” he says, and he means it as a compliment to no one. The lesson is the lesson of the monorail. When the smooth-talking stranger arrives in town with a song and a straw hat, the question is not whether the song is catchy. The question is why no other town in the country has bought what he is selling. If you are the only buyer in the world for a deal that sounds too good to be true, you are not getting a deal. You are getting played. Canadians, in his reading, have been getting played for 70 years and have developed an emotional attachment to the experience that they are only now beginning to question.

Maverix

By 2017, John could see that the problem he had originally come to OMERS to solve had been only partially addressed. He had built a firm capable of getting Canadian companies from $0 to $10 million of revenue. He had built a firm capable of getting them from $10 million to $100 million. But the third problem, the one that was now staring him in the face as Shopify and its peers approached the next stage of scale, was the problem of getting Canadian technology companies from a $100 million of revenue to $1 billion and beyond. There was, again, no one in Canada writing the cheques required to do that work. Whatever was going to fill the gap was going to need to be built from scratch.

He went to OMERS leadership and proposed building exactly that, a growth private equity practice within the OMERS umbrella that would write $100 million cheques into Canadian technology companies at the late growth stage. He got approval, and started hiring a team. And then, almost immediately, the institutional politics that he had managed to avoid for the better part of a decade began to bite. The traditional private equity team at OMERS started asking why he was building something so close to its territory. The venture team, which had been his own creation, started asking why he was scaling above his lane. He found himself squeezed between two larger and more entrenched groups inside the same institution, and as 2018 wore on it became increasingly clear to him that the political ceiling he was operating under was not going to lift. The firm he was trying to build inside OMERS was, in essence, the firm he wanted to build period. At the end of 2018 he left.

In a single week in early January of 2019 he wrote the white paper that would become the founding document of Maverix Private Equity. The thesis was Zeihan-shaped from beginning to end. The world was about to enter a great unwinding. Globalization, as it had existed since 1945, would fragment. The countries that controlled the conditions of their own existence would do well; the countries that had outsourced those conditions to others would not. Maverix would invest in the industries Canada needed to control, defined as health and wellness, financial services, transportation and logistics, live-work-play-and-learn, and retail. Technology was no longer a sector. Technology was something every traditional industry was about to be either rebuilt around or replaced by. The right play was not to invest in technology companies. The right play was to invest in the operators of traditional industries who understood how to weaponize technology against incumbents. Viral Nation, the Toronto-based social media influencer platform that became the firm’s first investment, was the proof point. Marketing business at the core, technology business in the architecture, growing globally, headquartered in Canada.

The name of the firm came from a small accident of the early days of the fundraise. The first three people who returned his calls after he announced his departure from OMERS each asked, in slightly different ways, how he had managed to last inside such a large institution. They asked it in a tone that suggested they were surprised he had stayed 8 years. By the third call he had figured out what they were getting at. He was, in their telling, a maverick, and mavericks did not last inside large institutions. He hung up the phone, looked across the room at the lawyer he was working with on the fund’s incorporation, and told her they had just figured out the name of the firm.

He went out to fundraise in early 2019 with a target of US$500M. He landed it within three months. Three months later, the lead investor reneged on his commitment for reasons that John found infuriating and that he was never quite able to forgive. He had to start the second half of the raise from scratch. By October of 2019 he had two new co-leads in place and a closing date set for the spring of 2020. The closing date was March of 2020. The pandemic shut down the process the same week. He waited. He waited some more. The pension fund partners began to come back to the table in late summer of 2020. They set a new closing date for October. He breathed, and went back to his bicycle.

Pickering Town Line

The bicycle, for John, was both a hobby and a refuge. He had been a serious cyclist for most of a decade, riding several days a week, training his upper body in a way that most cyclists do not because he liked to climb hills and he wanted the power. He did not, as a general rule, like to ride inside the city of Toronto. He thought Toronto was a dangerous city for cyclists, and he had long ago developed a route that allowed him to slip out of the urban core as efficiently as possible and into the safer roads of York Region. During the pandemic, with the rest of his social life on hold and his fundraising in suspended animation, the cycling had grown into a kind of midweek meditation. He was in the best physical shape of his life. He weighed about 200 pounds. He had powerful legs and an unusually strong upper body. He rode 4 or 5 days a week. The route from his house in the Beaches to the small town of Stouffville, north-east of the city, was about a 100 kilometres in a loop, and on Wednesdays in the summer of 2020 he was riding it almost without fail.

On Wednesday, the second of September, 2020, it was a beautiful day. He left his house in the late morning. He felt, by his own account, on top of the world. The fund was about to close. The political winds were shifting in his favour. The investors were lined up. He pedaled north out of the city along his usual back-road route, came out into the country, and turned east on Pickering Town Line, a quiet two-lane road that ran straight through the farmland a few kilometres south of Stouffville. He had ridden this stretch a hundred times. He was about 45 kilometres into the ride, 5 kilometres from his customary coffee stop in the small downtown of Stouffville.

What he heard, very clearly, with no warning, was the squealing of air brakes from a tractor-trailer directly behind his left ear.

His first reaction, in the half-second that he had to react, was anger. He assumed, on the basis of a hundred prior incidents, that some idiot truck driver was riding right on his back tire to scare him, and he was preparing to get off the bike and confront the man. But, he did not have time to confront anyone. The truck, a flatbed eighteen-wheeler whose driver had apparently been looking at his phone, hit him in the back at roughly 90 kilometres an hour. The bike was destroyed instantly. Two metal road signs anchored in concrete six feet off the road were taken cleanly out of the ground by the trailer as it jackknifed. He flew through the air, landing on the right side of his body. He did not break a single bone in his arms or in his legs, but broke, in multiple places, every single rib on the right side of his chest, in a pattern that doctors call a flail, in which a section of the rib cage separates from the rest of the chest and moves in the opposite direction during breathing, which is supposed to make survival, on its own, almost impossible. He shattered his pelvis into six pieces in what is called an open book fracture. He shattered the T12 vertebra in his lower spine, severing the connection to the lower half of his body. He tore his liver and his spleen, devascularised one of his kidneys, with both of his lungs collapsing. He lost approximately 50% of the blood in his body onto the asphalt.

He should have died on impact. Every doctor who later read his file said the same thing. He should have died on impact, and if he had not died on impact he should have died of suffocation in the first few minutes from the flail chest, and if he had somehow survived that he should have died of blood loss before any ambulance could possibly reach him. None of these things happened. He woke up on the road, two minutes after the collision, with traffic stopped around him and bystanders certain that he was already dead, and the first thing he did, with the practiced reflex of a man who had been in cycling accidents before, was to wiggle his fingers and his toes. The fingers worked, but the toes did not. He understood, in that moment, that he was paralysed. The second thing that came into his mind was that his life, in the broader sense, had been a good one, and that it would be acceptable to die on this road, in the country, doing something he loved. The third thing that came into his mind was the faces of his two children. His son was 14, and his daughter 11. He thought, “there is no way I am going to let my kids find out their dad died like this.” He tried to push himself up.

The ambulance ride to Sunnybrook Hospital in north Toronto took about 25 minutes. The paramedics could not airlift him because they had less than an hour to get him to a trauma surgeon and ground transport was, perversely, faster. He was conscious for the entire ride, on a stretcher that was not designed to be a bed, with his back broken and his spinal cord exposed, feeling every single bump in the pavement. He apologised in advance to the paramedic riding with him for the language he was about to use. The paramedic told him to swear away. He swore through the entire ride. The last thing he saw, as they wheeled him through the doors of the Sunnybrook emergency department, was the sign for the trauma unit. Then they put a mask on him and he was gone.

The surgery that followed took 15 hours and could not be performed for the first 36 hours after his arrival, because the doctors believed that a third major trauma on top of the injuries he had sustained on the road and on impact would kill him on the operating table. They waited, and told his wife, Carryn, that they wanted him to die peacefully if he was going to die. He did not die, and after 36 hours they took him into the operating theatre and worked on him through the night, rotating his body on a kind of surgical rotisserie to repair the front and back of his pelvis with screws and titanium rods and to fuse the lower part of his spine to his pelvis. He was in a medically induced coma for 6 days afterwards.

While he was in the coma the strangest thing in this entire story happened. Carryn had brought a pair of headphones to the hospital and was playing recordings into his ears, on the theory that some part of him might hear and might find the sound comforting. The recordings included Andrea Bocelli, the great Italian tenor whose voice, in the world of John’s parents, was the closest thing to a national treasure. John, lucid in some half-conscious way that has never been fully understood, heard the music. He saw white light, the way people in near-death experiences are reported to see white light. He saw, in the white light, what he understood to be a stage, and he heard Bocelli singing and an audience clapping. He understood that he was being welcomed into heaven by the angels, that the angels were Italian, and that they were applauding their countryman. The boy from Jane and Sheppard was getting a fitting send-off. He was sad about leaving his children. He was at peace about everything else. When he woke from the coma several days later and told his wife about the vision, half-laughing at himself, she stared at him and told him about the headphones.

Eight days after the accident, when the breathing tube had been removed and he could speak, the doctors came in to deliver the formal diagnosis. ASIAA Complete. The most severe category of spinal cord injury. He would never walk again. He cried for somewhere between 15 to 20 seconds, then stopped crying, and looked at the doctors and at his wife and said, in the single most characteristic sentence of his entire life, that they were not going to tell him what he was or was not going to do. He was going to walk again, and would not be told otherwise.

Six weeks into his four-month stay at the rehabilitation hospital, lying flat on his back, John picked up a telephone and resumed his fundraising for Maverix. He called the pension funds, family offices, and the institutional investors who had been on track to close the fund in October. They picked up the phone, in many cases, having not yet heard about what had happened to him, and they would chat for 10 or 15 minutes about the markets and the deal pipeline before some of them would pause and ask, with a confused tone, where exactly he was calling from. He would tell them. They would ask if he was on his back in a hospital bed. He would say yes. They would tell him to please, please, take care of his health, and that they were not going anywhere, and that the fund would be there when he was ready. Every single one of them honoured that commitment. The fund closed in April of 2021, at $500M, the largest first-time fund ever raised in Canadian history, raised in significant part from a hospital bed by a man who had been declared dead at the scene of his accident.

He went home for Christmas that year, on December 23rd of 2020, against the advice of his doctors. He sat at his kitchen table with his wife and his two children and explained to them, gently, that their father had nearly died, that the doctors had not expected him to survive, and that there was a version of the story in which he was not there for that dinner. The children, who had lived through a long parade of his prior cycling accidents and had become casually accustomed to their father bouncing back from catastrophe, looked at him and said that they had never thought he was going to die. They had seen him through too many incidents already. He was the cockroach. “Could someone,” his son asked, “pass the potatoes.”

In the years since, he has put in 20 hours a week of physiotherapy, every week, without exception, for 5 consecutive years and counting. He has cycled more than 3,000 kilometres a year on a recumbent bicycle that allows him to clip his feet into the pedals. He has walked, with a walker and then with two poles, distances that the original team of doctors at Sunnybrook believed were medically impossible and have, since seeing him do it, openly admitted they do not understand. He has raised, through his cycling team Les Domestique, more than $25 million for spinal cord and brain injury research. He has joined the board of a Canadian biotechnology company called NervGen Pharma, currently in late-stage human trials of what may turn out to be one of the most significant treatments for spinal cord injury, multiple sclerosis, and several other diseases of the central nervous system in the modern history of medicine.

Today

It is 2026. He is close to 40 years in this work, and he is not getting younger. Maverix has made its first investments and is doing exactly what he built it to do, writing large cheques into operators of traditional industries who understand technology, in the sectors he believes Canada most urgently needs to control. The thesis he started writing in 2019 has, in the years since, been violently confirmed. Globalization is fragmenting, supply chains are shortening, and energy markets are regionalizing. The American posture toward Canada has, against every expectation of every Canadian politician of the post-1945 era, turned actively hostile. The Trump administration is treating Canada not as a partner inside the fortress but as a potential vassal outside of it. The Fortress North America scenario John and Zeihan had originally expected, in which Canada would be the junior partner inside the American security and economic perimeter, is no longer the base case. The base case is something more dangerous, and the timeline for building Canadian sovereign capability across food, energy, defence, money, and data has compressed from decades to years.

He has views on what should be done. He has them in detail and he will share them at length to anyone who asks. The pension funds, which he has been hammering for 20 years, remain his single biggest target. Their argument that they should invest globally rather than domestically is, in his reading, a category error. The whole reason they exist on the favourable taxation terms they enjoy is that they are creatures of Canadian public policy, sustained by tax expenditures borne by Canadian taxpayers. Their refusal to allocate even a small fraction of their assets to Canadian risk-based investment, while they happily fund SpaceX and pass on Kepler, is not just bad investing. It is a betrayal of the institutional logic that allowed them to exist in the first place.

At the same time, the federal government continues to spend the vast majority of its technology budget with American hyperscalers, on infrastructure subject to the U.S. CLOUD Act, which means the United States government can compel the disclosure of Canadian government data without notifying the Canadian government. He warned about this for years before any politician understood what he was talking about. In June of 2025 the president of Microsoft, Brad Smith, said publicly in France that Microsoft would sue the U.S. government if it tried to compel data of European customers under the Act. Smith confirmed, in effect, that the Act worked the way John had been saying it worked. John forwarded the clip to every Canadian official he had ever briefed. They had not believed him before, but had to believe him now.

Currently. the artificial intelligence question is the sovereignty question of the moment. In the rule-of-three logic he believes governs the digital economy, three players will eventually capture most of the value in any given category. There will be three foundational large language model providers, three hyperscaler clouds and three dominant chipmakers. The Canadian question is not whether to regulate artificial intelligence, but whether to be a price taker on infrastructure provided by foreign incumbents, or to identify the choke points in the AI supply chain where Canada has genuine native capability and to invest aggressively in those choke points. Photonics is one such choke point. Canada is one of three countries in the world capable of producing the photonic chipsets that drive the next generation of high-throughput computing and satellite data transmission. The other two are Taiwan and the Netherlands. Most Canadians, including most of the Canadian government, do not know this and he is trying to fix that. Quantum computing, in the photonic variant pioneered by Xanadu, is another choke point. Defence technologies, where Maverix has been building a thesis around dual-use capability, is a third. Energy, where Canada’s geographic endowment is unmatched and where the country has spent 50 years systematically refusing to capture the value of its own resources, is a fourth. The list is short, the time is short, and the political will is, for the first time in his career, beginning to materialize in pieces. Mark Carney’s senior team, he allows, is starting to understand the problem. There is, he says, 99% more to go.

What he does not allow, in any conversation, is the line that Canadian sovereignty is somehow incompatible with foreign investment, or that the argument is protectionist. Foreign capital is welcome, with the condition that the rules of engagement are written by Canadians for Canadian benefit, and that the structures of the investments preserve, rather than transfer, the ownership of intellectual property. The battery plant deal that committed $13 billion of Canadian taxpayer subsidy to a foreign-owned facility in exchange for a thousand jobs is, in his reading, exactly the wrong template. Such an amount deployed into Canadian-owned innovation, even at the very high failure rates of venture capital, would create perpetual wealth creation machines, but deploying into a foreign-owned battery plant creates a thousand jobs and a press release. The choice between the two templates is the choice between sovereignty and dependence. Canadians have been choosing dependence on autopilot for half a century, and he is trying, with the time he has left, to persuade them to make a different choice.

He worries, more than he ever expected to worry, about his own children. They are 17 and 20 now. They are starting to talk, the way ambitious young Canadians have been talking for two generations, about going to the United States. He does not want to stop them. He wants them to see the world, study where they want to study. What he wants is for them to come back. He has never, until recently, worried that they would not. He worries now.

He has been knighted by the President of Italy, and holds the title Cavaliere. He has one foot in Italy, and is connected enough to its institutions that he could, if he chose, build a life there. He does not choose to. He was born in Canada, and he will die in Canada. The argument he has been making for 34 years is, in the end, an argument about whether the country he was born in will still exist, in any meaningful sense, when his children’s children are old enough to have the conversation about whether to stay.

His extended Italian family went into the trades because every other door was closed. They built businesses because they had no choice. They lived in a cocoon at Jane and Sheppard because the cocoon was the only place that was safe. Every generation since has had more options than the last, and the boy who once spent his entire adolescence trying to escape the cocoon is now, 60, trying to build a national version of it, a structure of capital and rules and institutions that gives Canadian founders the same protected ground his parents’ generation gave him without ever knowing they were giving it. He would not call it that. He would call it sovereignty. They would have called it home.

The Lyle Lanleys keep arriving in Springfield with their songs and their straw hats. Most of the time they win. Sometimes, when somebody in the crowd is paying close enough attention to see the thing for what it is and is willing to take the personal cost of standing up and saying so, they lose.

He intends to be in the room when they come.

He is, as he was always going to be, a maverick.

In the late 1990s, in a French-language elementary school on the western edge of the Montérégie, Andrew Lockhead was a gangly boy who never seemed to know where to put his hands. In a classroom full of Tremblays and Bélangers, his name didn’t even sound like a name. He was the mid-year transfer kid with hair pressed flat and glasses so thick they magnified his eyes into wide, unblinking saucers. Fond de bouteille, he’d call them later, coke-bottle lenses. In his first months at the school, he broke them so often that his mother kept a spare pair in the glove compartment, a permanent fixture for the drives she made every few days to the optician.

But before that, there had been Montreal, where Andrew was born in 1987. His father, Pierre Lockhead, held two jobs. One was as a cashier at the Montreal Blue Bonnets Raceway, a horse racing track, where he sat behind the glass of a betting window for full shifts while men pushed 20 cents under the slot, asking for tickets on horses they had picked out of dog-eared programs. His second job was as a mechanic on the floor of a car shop somewhere else in the city, working for someone else. His mother also worked, which meant the daycare Andrew shared with his younger brother sat in the basement of a Catholic church.

The decision to leave Montreal came from both parents, but for different reasons. His mother had begun to worry about her sons’ French. The elementary school they attended in their Montreal neighbourhood had only two children in Andrew’s class who spoke French at home. While she liked the exposure to other languages, she decided her boys were not going to grow up in a Quebec where they could not hold a conversation in their mother tongue. His father had decided that the racetrack window and the back room of someone else’s garage were the last jobs he was going to work for a boss. He had been thinking about it for years. He and a partner bought into a franchise of a Quebec car window repair chain, signed the paperwork on a unit in the West Island, and packed up the apartment. The family moved to Pincourt in the autumn of 1993.

Window Repair Shop

Montreal to Pincourt is a short 20-minute drive on the highway if traffic is moving. In the early ‘90s, it was suburban, quiet, and still small enough that the new subdivisions ended in farm fields and the elementary schools sat on lots big enough to play soccer behind. Andrew’s father opened the first garage and it did really well. Within a few years, he had opened a second one, then a third, then a fourth, scattered across the West Island and the eastern side of the island of Montreal.

The family’s life began to change in ways that were measurable at the level of materialism. Andrew’s younger brother went to private school, his father bought a car that started on the first turn of the key, and they began going on vacation. His mother stopped working outside the home and became the headhome, the role of a Quebec stay-at-home mother who had effectively been promoted into the operations manager of the household. She drove the boys to school in the mornings, packed their snacks, made dinner, and drove Andrew, every two or three days, to the optician.

The trips to the optician started almost as soon as Andrew arrived at the new school. The pair of glasses he was wearing the day he started would rarely be the pair he was wearing the next week. The frame would come back bent at an angle his father could not straighten by hand, with a lens cracked, or with the bridge snapped clean in half. His mother kept a spare pair of glasses in the glove compartment of her car for the inevitable moment, which arrived several times a week, when she had to drive him from school back to the strip mall off the main road. The optician there had come to know the family by name. He performed the repairs, usually temporary ones that would hold the frames together until Thursday or Friday, at which point Andrew would be back in the chair waiting for the next adjustment.

The protector in the family was Andrew’s brother, who was two years younger and built closer to the ground. He could absorb a hit and keep walking in a way Andrew never could. When Andrew was 9 and coming home with broken glasses every other day, his brother was 7, and he was usually the one who stepped in and stopped the fights before Andrew even had to be in them. He is a firefighter in Montreal today, continuing to protect the innocent.

The fights eased toward the end of Andrew’s second year at the new school, and the easing happened without any single triggering event Andrew can remember. There is no scene in his telling where he stood up to the lead bully, won a fight, or said the right thing at the right moment. The fights simply slowed down as other targets emerged in the schoolyard. His glasses would still break and his mother still drove him to the optician, but the frequency went from several times a week to several times a month, which was by the standards of the household at that point, an enormous improvement.

He was around 10 when his father first brought him out to the garage to work on a Saturday morning. His brother came along too. The Saturday-morning trips would continue, in one form or another, for most of the next decade of Andrew’s childhood. However, there was no pay attached. His father had decided early on that a trade was not something to be learned for a wage. The boys were there to learn, and the money, if it came, would come later. They were not paid for the first four years.

The garage itself was a single-story franchise unit on the edge of a service road off Highways 20 and 40, with two service bays at the back and a small customer waiting area at the front. The smell of the place, the same in every garage Andrew’s father would later open, was the smell of motor oil, burnt drip-coffee, and the rubber of new tires waiting in the corner. There was a counter at the front, and behind that was a glass partition through which the customer could see the bays where the cars sat on the lifts. Andrew and his brother spent most of their Saturdays in the front. They stood behind the counter, handling customers, taking phone calls, writing up estimates, and walking buyers through the catalog of windshields and radios.

The back part of the garage, unfortunately, was something his father could not seem to master. The mechanics his father employed in those years were, by Andrew’s account, poorly paid and poorly trained. He had pieced together this picture from what his father said about them at home and from what he could see on Saturday mornings, when the previous day’s unfinished work was still sitting in the bays. They would routinely fail to show up on Fridays after spending Thursday night drinking the previous week’s wages, shouting at customers when his father stepped away from the front counter, and taking longer on simple repairs than the work actually required.

The lesson Andrew took from those Saturdays was that the front and the back of a small business were not the same operation. The front was where leadership lived and where outcomes could be shaped. The back was where culture either worked for the owner or against, and in his father’s particular industry, the back was a labour pool his father could not change no matter what he did at the counter. By the time he was a teenager, he had decided that whatever business he eventually built would not have a back like that.

StudentSphere

The decision to leave the French school system for an English-language CEGEP was made in a series of conversations with his parents and a career counselor over the months before he turned 18. His father wanted him to come into the family business once he finished school. The counselor told him that the best thing he could do before joining anything was to get a credential that would not close any doors for him later. His father agreed. The question then became which institution to attend, and Andrew chose John Abbott College, the English-language CEGEP in the West Island. The practical reason was that he could not speak English and was going to need it for any business career the counselor wanted him to keep the door open for. The less practical reason was that John Abbott was a place where almost nobody he had grown up with would be enrolled. Out of his entire French high school, only one other student had decided to go. The idea of arriving somewhere where nobody knew him was the kind of fresh start his elementary-school self would have wept with relief to see coming.

The glasses came off the summer before he started at John Abbott. The standard medical advice for laser-correction surgery in those years was to wait until the patient was at least in his 20s, but Andrew’s prescription was bad enough that the surgeon was willing to take him on early, and his father had a contact through the garage who could get him into a private clinic. His father wrote a check for $12,000 and told his son to be at the clinic in the morning. The procedure worked. Andrew arrived at John Abbott orientation a few weeks later without the glasses, presenting a face that the kids he had grown up with would not have immediately recognized at a distance.

His first English class was humanities. He sat down at his desk. The teacher began to speak, and Andrew, who had not spoken English to anyone in his life beyond a few phrases of broken classroom translation, watched the teacher’s mouth move and understood almost nothing. The other students were opening their textbooks and writing notes, but he had no idea what they were writing. He sat through the rest of the class with his hands flat on the desk in front of him and asked himself, in French, what he had just done to his own life.

He waited for the bell to ring and the room to empty, then walked up to the teacher and explained, in French, that he could not follow. The teacher did not speak French. They communicated through hand gestures and broken phrases, eventually arriving at an arrangement: Andrew would come find the professor before every exam and during office hours to work through the material together, line by line, until he understood enough of it to write the test. He did this with every class he took for the first semester, and eventually graduated on the dean’s list.

He applied to every business school in Montreal after John Abbott. McGill rejected him. The John Molson School of Business at Concordia rejected him. HEC rejected him. The only school that accepted his application was ESG-UQAM, the management school at the Université du Québec à Montréal. Andrew was missing a math credit from his time at John Abbott, and ESG-UQAM was the only school willing to admit him, on the condition that he take the missing course as a remedial requirement during his first year. He took the offer. The commute from his parents’ house in Pincourt to the downtown UQAM campus was a 1.5 hours each way on the highway and the metro, which meant that on days he had a full class load, he was on the road for three hours and in the classroom for three more. He kept his grades up. He also began looking with an intensity for something to do with the hours he was spending in the city that were not the hours he was spending in the classroom.

The first day of his first frosh week at UQAM, at a beach party the business student association had organized on an island in the Saint Lawrence, he saw a man standing at the grill flipping burgers with a walkie-talkie clipped to his belt, clearly the person in charge. Andrew watched the man for some time. Then he walked up to him, with a confidence that had not previously appeared in any other domain of his life, and told him plainly that he was going to do his job in two years.

Three years later, Andrew was the president of the student association of his faculty, an organization with roughly 12,000 members, a budget of $2.5 million, a coffee shop, a career center, and an annual calendar of events that included regularly booking Metropolis, a Montreal venue with a capacity of 2,500. He took over the events portfolio in his second year, before he took the presidency, and the first major event he ran almost broke him. A week before the show, which he had personally guaranteed against the association’s budget, they had sold almost no tickets. He went home on the Sunday night before the event certain that on Friday, he was going to be the youngest student association officer in the history of UQAM to bankrupt his union. Andrew worked 15+ hours every day of that week promoting the event, until he could sell every ticket he possibly had. The show went off as the biggest party UQAM ran that academic year. Andrew sat in the dressing room afterward with the production crew and understood for the first time what kind of work he wanted to spend his life doing.

He also got dragged through two strikes during his time at the association. The bigger of the two grew out of the 2012 Quebec tuition fight, in which the provincial government had proposed an increase to post-secondary fees and the Quebec federation of student unions, the FEUQ, had called the universities into the streets. The students Andrew represented at his business faculty were mostly opposed to the strike. They were going to pay their own way through school, so the increase did not register to them as an existential threat, and most preferred to keep going to class rather than spend a semester on the picket line. The other faculties at UQAM saw it differently. Strikers from sociology and political science came into the business school building and tried to disrupt classes. They formed lines outside the doors of his faculty’s lecture halls. The police got involved. Andrew, 24 years old, spent that semester in a series of negotiations he had not expected to be conducting, with student leaders from the other faculties, with the university administration, and with his own membership. He treats the experience now as the closest thing to running a real business that an unpaid undergraduate job could have given him. He had to manage stakeholders whose interests did not align. He had to figure out what each of them actually wanted, which was sometimes power, sometimes recognition, and sometimes a line on a resume they could use to apply for jobs later, and he had to lead them on that basis.

The pain point that produced his first real company was something he had been living with every week of his presidency. To sell tickets to events his association ran, he had to stand in the corridors of UQAM at 8 a.m. with a cash box, calling out to students walking past, taking bills, making change, and counting receipts at the end of the day. Ticketmaster did not serve the student market in 2012. Eventbrite had launched but had not made it deeply into Quebec. Andrew and a friend from university decided to build the platform themselves. They called it Sphère Étudiante, also known as StudentSphere. They put in $5,000 each and received a small grant from Fondation Montréal Inc, the foundation now known as Bonjour Startup Montréal. With the starting capital, they paid an outside Montreal development shop $40,000 to build the first version of the product.

Outsourcing the technology is the decision Andrew now points to as the most expensive learning of his early years. The platform shipped and worked, but when they wanted to add features the next year, the same shop quoted them $120,000. By the third year, the architecture they had paid an outside vendor to build had reached the point where it could no longer be extended. Whatever ambitions Andrew and his co-founders had for the company were constrained by a technology stack they had bought but did not own. StudentSphere continued for ten years as Andrew and his co-founders eventually moved themselves off the payroll, hired a General Manager, and reduced their own involvement to occasional board calls. The company was sold in 2022 to a UK firm that wanted the technology for a North American expansion, for an amount that came out to less than $100,000 for all the founders, on roughly a decade of work.

Stay22

In the summer of 2014, while continuing to work on StudentSphere, Andrew attended Startupfest, the annual conference held in the Old Port of Montreal where the Canadian early-stage technology ecosystem assembles for three days every July. Andrew was sitting in the audience watching a young engineer named Hamed pitch a B2C travel company. Hamed had been kicked out of Dawson College a few years earlier for hacking the school’s system, and he was building a product designed to take on Expedia, Booking, and Kayak directly. Andrew was impressed by Hamed, but he was unimpressed by the strategy. He introduced himself after the pitch and told him, to his face, that he was wrong about the whole premise of the company. Going B2C against the incumbents in travel, who spent more than $10 billion a year on marketing budgets between them, was not a winnable fight no matter how good the product was. The right way to compete in travel was to embed inside the distribution channels where people were already making travel-adjacent decisions, the platforms where they were already buying tickets to events.

They reconnected a few years later. They tested working together for a month, selling a few event integrations, and then applied to Travelport Labs, the corporate accelerator run by the global distribution system Travelport. Travelport accepts only five companies per year. They got in and moved to Denver for four months during the winter of 2016 into 2017. Andrew, who was so broke that he had convinced himself he hated traveling, got a crash course in an industry he had previously known nothing about. He learned the suppliers, the global distribution systems, the commission structures, and the way affiliate revenue flowed from a booking back through three or four layers of intermediaries. Travelport invested US$35,000 for 8% of the company. While that may seem expensive, the founders did not negotiate, as the education was worth more than the check.

The next accelerator they entered was FounderFuel, the Montreal-based program Andrew had previously applied to with StudentSphere and been rejected from. With Stay22, he and Hamed made it in. The terms were $200,000 for 5% of the company, paid in two tranches across a 12-week program. The team grew from two to six during the cohort, with two new hires from Hamed’s network and two of Andrew’s former UQAM classmates. Andrew paid everyone, including himself, $1,000 a month. He was 29 at the time. He owed his parents $10,000 from previous loans he had not been able to repay, and he was living on ramen and the half of the rent his girlfriend (now fiancée) was paying for him. Somewhere in the middle of that year, on the theory that if Stay22 failed he was going to need a stable career, he applied to become an air traffic controller, and washed out of the training program.

FounderFuel ended in the autumn of 2017 with a demo day pitched in front of an audience of 2,000 people. Andrew and Hamed had been told by a FounderFuel mentor named Isaac that a seed-stage startup raises money on one of three things. The first was a powerful vision the founder could articulate well enough that an investor would buy the story before the numbers existed, and Andrew was not, by his own admission, a strong vision storyteller. The second was pedigree, the kind of resume that read “Stanford” or “former Google engineer” across the top of the deck. Hamed had been kicked out of Dawson College before he finished his degree, and Andrew had only barely finished his own at UQAM, which meant the second door was closed to them as well. The third was traction, and that was the only path left. They were the first team in the building every morning and the last team to leave every night, every working day for the full duration of the program.

They closed a seed round on the strength of the demo day numbers: $750,000 at a valuation of $3.75 million. The lead investor was Seven Gates Ventures, a fund out of Vancouver that Andrew chose partly because the firm had the kind of West Coast network he and Hamed lacked. Real Ventures, the Montreal-based seed fund, also came in. So did several angel investors who had found their way to Stay22 through Anges Québec, including one who had moved to Montreal from Italy and had no obvious reason to be writing checks to a Canadian travel startup, but wrote one anyway.

Ides of March

By February of 2020, Stay22 was doing $110,000 USD in monthly recurring revenue and the team had grown to 24 people. A new financing round had been negotiated, term sheets were signed, and several investors had already wired their commitments. Andrew was 33 years old. He had spent more than a decade getting to the point where he was about to fundraise his way into the version of his career that other founders in his cohort had been living for some years.

The pandemic arrived in the second week of March. Stay22 sat at the intersection of travel and events, the two industries the global economy was about to close completely, and revenue fell 90% in a matter of weeks. The cash that was supposed to fund the next 18 months of operations had only partly landed in the company’s bank account. The investors who had committed but had not yet wired their tranches now made the calculation that the company was about to go bankrupt, deciding they would prefer to keep the money rather than send it.

Andrew called them. The first calls were polite, and the investors told him the money was coming next week. The next week came, but the money did not. His follow-up emails went unanswered, his calls began going to voicemail, and by April, he could no longer reach them at all. One of them, who was already on the cap table as an angel investor, told Andrew, when he finally got the man on the line, to sue him if he wanted to but that he was not sending the money. Another sent a paper check with written instructions not to deposit it on the grounds that there were no funds in the account it was drawn against. Andrew kept the check on his desk. He still has it, sitting next to his monitor as a daily reminder of never celebrating until the money hits the bank.

He had to take legal action against the investors who had signed and refused to wire. The lawsuit was not, in any meaningful sense, a recovery exercise. As a director of the company, he had a fiduciary obligation to the investors who had wired, the ones who had stood by him, to demonstrate that he had pursued every available legal remedy against the ones who had not. If he failed to do so, those supporting investors would have had grounds to sue him personally for the duty he had not discharged. After filing and opening up a case, he moved on to the question of whether the company would continue to exist.

He cut the team from 24 people to 8. The conversation he comes back to when he tells the story of that period was not with any of the laid-off employees, but with his accountant, a man named David at a Montreal firm called Finalytics. David had been billing the company close to five figures a month for the work he did on the books, tax filings, and regulatory paperwork. In the spring of 2020, Andrew and his COO, Simon, sat across from David and delivered the news that the company could no longer pay. Andrew had no idea when, if ever, he would be able to clear the debt, but the work still had to be done because the company had no operations team left to do it. David could have stood up, walked out of the meeting, and quit on the spot, and Andrew would have had no leverage to keep him. Instead, David agreed to keep working without compensation for an undefined period, on the strength of a relationship he and Andrew had built over several years. Six months later, when revenue had begun to return, Andrew started paying David again. Andrew believes that if David had walked out of that meeting and quit, Stay22 may not have survived. Without accounting and daily cash flow projections, Simon wouldn’t have a single hair left.

The remaining investors told him to keep going. He sat down with what was left of the team in the spring of 2020 and ran a two-week brainstorming sprint. They built three minimum viable products. The first one failed within days, and the team shut it down before they had spent any meaningful time on it. It was a map-based product for nurses and doctors who needed accommodation near hospitals they had been temporarily assigned to during the early months of the pandemic. The Quebec government, which they would have needed as a partner to make the product function at scale, did not move at the speed a startup needs from its partners, and the idea died.

The second was a cross-sell product for e-commerce stores on Shopify, which offered travel-adjacent services to customers who had just purchased something from a Quebec apparel brand or wellness company, like Prana, Ciele, Poches & Fils, or Café Liégeois. The product worked in the sense that it generated revenue, but the unit economics did not scale, and the team eventually let it go.

The pivot that worked emerged from a parallel set of conversations Andrew had been having with Quebec content creators who had stopped traveling internationally and had begun rediscovering their own province. The travel bloggers wrote about the rediscovery first, the Instagram accounts came after them, and the YouTube channels arrived last. Stay22 had a map-based product designed for events, but the bloggers did not want a map. Andrew sat down with them in workshops he ran personally over the course of two weeks, while Hamed coded in the evening what they had discussed in the afternoon. Every morning, Andrew presented the wireframes to the creators and noted their reactions, and they iterated again. By the end of the two weeks, the team had their third minimum viable product. By the end of the second month, they had product-market fit. The creators wanted a way to monetize the content they were already writing about travel through affiliate links to hotels, restaurants, and activities that appeared in the posts they were already publishing, and they were happy to give a software company access to their websites in exchange for a share of the revenue. This is the version of Stay22 that exists today.

Almost exactly one year later, in March 2021, Andrew learned that Airbnb was shutting down its affiliate program. He saw the news through industry channels before Airbnb officially called him, and for a few weeks, he allowed himself to believe that Stay22, given the depth of its integration with the platform, would be exempt. The phone call came in late March. The Airbnb relationship manager told him that Stay22 had been a great partner and that the team appreciated the work Andrew had done on the API, but that Airbnb’s strategic direction no longer accommodated affiliate partnerships. The company would have 60 days to wind down the integration. Airbnb represented 80% of Stay22’s bookings.

He called his mother, who was in Florida with his father at the time. He told her what had happened, that this was the third near-death event the company had faced in 18 months, and that perhaps the universe was trying to tell him something he had so far refused to hear. His mother told him that nothing happens for nothing in this life, that this was a sign he was supposed to keep going, and that there was a reason for it that he could not yet see.

He pivoted Stay22’s supply layer onto Trivago, the German meta-search engine, and the first month under the new arrangement was extraordinary. The company touched almost $500,000 in revenue in a single month, more than any month it had ever recorded. Andrew began to allow himself to believe that the Airbnb cut had been the best thing that had ever happened to the company. Then Trivago called with a correction to the numbers from that first month. The relationship manager explained that Trivago’s accounting had been overpaying Stay22 for clicks at an unsustainable rate, and that the rate would need to come down significantly.

Andrew was burning $200,000 a month against a runway of approximately three months, and the data did not support keeping the company open. The rational decision, by every framework he had taught himself to apply, was to wind it down and start something else. Andrew did not see the situation in those frameworks anymore. Raising more cash wasn’t on the table, not with a 90-day runway, and certainly not after getting burned by their last investor. He was done looking for a rescue party. Time was the only resource he could not replenish, and starting another company would burn years he could not recover. He was 34 years old, so starting over was no longer something he could treat as a free option. The creator base he and Hamed had spent the previous year building had 90 days to generate enough revenue to save the company. If it did not, he would close Stay22 at the end of the runway and move on.

It delivered. By the middle of 2022, the company was running at roughly $1.5 million in monthly revenue. By 2023, it was profitable, and it has been profitable every month since, doubling in revenue every year for the past five years. The team grew from the post-pandemic skeleton of eight to roughly 125 employees, tripling the team size within two years. Andrew personally conducted the final interview for every hire. He had learned, from an earlier batch-hiring spree that had broken the company’s culture in ways he had spent two years repairing, that hiring more than 15% of a team per quarter was the most reliable way to destroy a small company’s culture. The one-employee-a-week cadence was his answer to that mistake.

Every March since the pandemic, the company has faced some kind of existential threat. March 2020 was COVID. March 2021 was Airbnb. March 2024 was a brutal Google algorithm update that stripped Stay22’s independent content-creator partners of 40% of their search traffic, forcing the company to pivot hard toward major digital publishers who still held sway on the open web. By 2025, a second paradigm shift arrived: the rise of LLMs and AI search began draining traffic away from text-based Google results entirely, sparking a massive exodus toward YouTube, Instagram, and social platforms where creators could build un-scrapable, direct relationships with their audiences. Each March, the question has been the same, and the answer has been to keep going. There’s an opportunity in every crisis that is faced, and at all times, a deep dive should be done to find that opportunity.

Today

In February 2026, Stay22 announced a US$122M minority growth investment from Summit Partners, the Boston-based growth equity firm. The deal was the company’s first institutional capital event since the seed round Andrew and Hamed had closed in 2020. By the time it priced, Stay22 was processing more than US$1 billion in annual transactions and working with more than 5,500 creators. On the morning of March 11, 2021, Andrew had been sitting in his office reading an email from Airbnb and asking himself whether the universe was trying to tell him to fold.

He runs the company today on a set of operating principles that have hardened over time. He implemented the Entrepreneurial Operating System in 2019. He runs the executive team on the framework Patrick Lencioni laid out in Five Dysfunctions of a Team. He recommends Annie Duke’s Quit: The Power of Knowing When to Walk Away to other founders because he believes the harder problem in entrepreneurship is not perseverance but the discipline of knowing when to stop.

The conversation that reset his understanding of his own role as chief executive happened within the past year. He had been trying for two years to be the kind of CEO he believed the company needed at its new scale, a more delegating, bottom-up version of himself, a man who built management systems and trusted his executives to run them. Then, his COO, Simon, asked for a meeting. He told Andrew plainly that the experiment was not working. He explained that Andrew was a wartime founder who naturally belonged inside the details of the business, and that he had been operating against his own nature for two years. He was miserable in the process, and Simon told him to stop fooling himself. Andrew took the feedback, leaned back into the active work, and the company immediately started moving faster.

He lives now in Laval, the town next to the one he grew up in. His daughter will turn 3 and his son 6 this summer. He is home only one evening of the work week, and sometimes none at all. But he makes sure he is home every weekend without exception, and on those days, he does not look at his phone while the children are awake. His brother remains a firefighter in Montreal. His father retired at 53, after selling his chain of window repair shops and reaching financial independence. Today, his parents spend six months of every year in Florida among the other Quebec snowbirds.

Andrew turns 39 this summer. He previously said that he hoped to surpass his father’s success by the age his father retired. With the Summit Partners deal, he has already surpassed his father in raw dollar terms.

But the true difference between the two men is no longer the money. It is that his father retired the moment he had enough, and Andrew has not yet figured out what enough looks like. Maybe in the next few years he will reach the point his father reached. Maybe not. Who knows?

On March 19, 2026, a threat actor with the handle TeamPCP exploited a single misconfigured GitHub Actions workflow inside Aqua Security’s Trivy repository, one of the most trusted open-source vulnerability scanners in the world. They compromised a service account, force-pushed malicious code to 76 of 77 version tags, and quietly embedded a credential stealer into a tool that security teams around the world were actively using to protect themselves.

The payload harvested SSH keys, cloud tokens, Kubernetes secrets, and npm credentials from CI/CD environments. Those stolen tokens fueled CanisterWorm, a self-propagating worm that cascaded through 66+ npm packages using blockchain-based command infrastructure that couldn’t be conventionally taken down. Within eight days, the same campaign had compromised GitHub Actions, Docker Hub, npm, PyPI, and the VS Code extension marketplace. One misconfiguration. Five ecosystems. An estimated 300 GB of exfiltrated data. Over 500,000 stolen credentials.

The attacker didn’t write a zero-day. They didn’t break encryption. They turned a trusted security tool into a weapon, and most organizations never saw it coming.

This Isn’t Your Father’s Supply Chain Attack

The industry has been talking about software supply chain risk since SolarWinds in 2020. But something fundamental has shifted in the last twelve months.

The attack surface is no longer human-scale.

The average application ships with over 1,100 open-source components. Most of those were chosen by nobody on your team: they’re transitive dependencies, packages that your packages depend on. And increasingly, even first-order dependency decisions are no longer made by humans.

AI coding agents (Cursor, GitHub Copilot Workspace, Claude Code, and their descendants) are now writing code, selecting libraries, and opening pull requests without a human ever touching the keyboard. A study of over 117,000 dependency changes found that AI agents select known-vulnerable dependency versions 50% more often than humans. They do this at speeds that compress the security review window to essentially zero.

The same week as TeamPCP, attackers also compromised Axios (the HTTP library downloaded over 100 million times a week) by adding a malicious dependency that ran a remote access trojan on install, then self-destructed before anyone noticed. The industry average time to detect a supply chain breach is 267 days. On 135 monitored endpoints, that malware executed and phoned home to the attacker’s server within 89 seconds of install.

As a16z put it bluntly in their April 2026 analysis: “We are building a world where machines write the code, machines choose the dependencies, and machines ship the updates.” If security doesn’t keep pace, the AI agents are cooked.

The Three Attack Surfaces Nobody Is Defending

There are now three distinct attack surfaces that most security teams have almost no visibility into.

The developer machine. Developer machines are running MCP servers, AI models, IDE extensions, and browser plugins that directly influence what code gets written and committed. Credentials accumulate across dotfiles, .env files, and environment variables. Security teams typically have no real-time inventory of what’s running.

The coding agent. AI coding agents don’t just suggest code: they act. They install dependencies, invoke external tools via MCP servers, execute builds, and push commits. Most security policies were written for humans. There is no category in your existing tooling for “approve this MCP server plugin before the agent can use it.” Attackers are beginning to exploit this gap explicitly.

AI-generated code. LLMs regularly invent package names that don’t exist. Nearly 20% of AI-recommended packages are fabrications, and attackers register these hallucinated names in advance with malicious payloads. The technique is called “slopsquatting.” One researcher uploaded a dummy package with a commonly hallucinated name and watched it accumulate 30,000 downloads, largely from AI-driven workflows, in weeks.

Traditional security tooling is blind to all three surfaces. Most software composition analysis tools work by checking dependencies against CVE databases. But a newly planted backdoor doesn’t have a CVE. Running npm audit on the compromised Axios version returned a clean bill of health because the malware had already self-destructed.

Securing AI Development at the Source

This is the problem that Montreal-based Boost Security is built to solve. And why we at Amiral Ventures are proud to have backed them.

Boost Security’s core insight is deceptively simple: the right place to stop a supply chain attack is not the CI/CD pipeline gate. By the time code reaches your scanner, credentials may already be exfiltrated and your developer machine may already be compromised. You need to move protection upstream: to the moment a prompt is sent to a coding agent, to the moment a dependency is suggested, to the moment a plugin is installed.

Their newly launched Developer Endpoint Security platform gives security teams centralized visibility and governance across the full AI development workflow:

• Developer Endpoint Visibility: A real-time inventory of every coding agent, MCP server, AI model, IDE extension, and package running across your developer fleet, the exact visibility gap that made TeamPCP possible.

• Coding Agent Safety: Governance controls ensuring agents only run with approved MCP servers and plugins, with configuration drift flagged before it becomes an incident.

• Supply Chain Security: Behavioral analysis of packages and extensions: evaluating what code actually does, not just checking it against a CVE list.

• Secure Agentic Code Generation: Guardrails embedded into the coding agent workflow so generated code follows organizational secure coding guidelines before being committed.

• Data Leakage Prevention: Outbound prompt scanning to detect and mask credentials and API keys before they reach external LLMs.

“AI coding agents are fundamentally changing how software gets built, but security has largely remained focused on scanning code after the fact,” said CEO and Founder Zaid Al Hamami. “Developer Endpoint Security moves protection upstream. It secures the developer machine, governs the coding agent, and ensures safer code is generated from the start.”

Why We Invested

A few things stood out when we evaluated Boost Security at Amiral.

The team. Zaid co-founded the company alongside Rajiv Sinha, who built Oracle’s first Application Security program and later worked at Cigital, the leading North American AppSec consulting firm. This is Zaid's second software security startup. His first, Immunio, built the first RASP technology, and was acquired by TrendMicro. Between them, two decades of seeing every era of software security from the inside. Their VP of Security Research, François Proulx, is one of the founders of NorthSec and a veteran AppSec researcher who has discovered 0-days in Terraform providers, AWS Helm Charts, and major GitHub Actions; he’s a recognized voice in the supply chain security community long before most people knew it was a category.

The timing. The threat landscape shifted materially in the past 12 months. AI coding agents have moved from prototype to production infrastructure at thousands of organizations. The attack surface they create, developer machines running MCP servers, agents making autonomous dependency decisions, LLMs hallucinating package names, didn’t meaningfully exist two years ago. Boost Security is building for the world that actually exists now.

The market. After SolarWinds, Log4Shell, XZ Utils, TeamPCP, and the Axios attack, the CISO conversation has shifted from “should we care about this?” to “how do we actually defend against it?” The specific category of AI-native developer security (securing the coding agent workflow, not just the code it produces) is still early. There are very few credible solutions addressing all three attack surfaces described above. Boost Security is one of them.

What This Means for You

If you’re a CISO or AppSec leader, three questions worth asking your team today:

1. Do you have a real-time inventory of every AI tool, MCP server, and IDE extension running on your developer machines?

2. Do your security policies govern what AI coding agents can connect to and what packages they can install?

3. Are you checking packages for behavioral signals, or just CVE matches?

If the answers are uncertain, this is exactly the gap Boost Security is built to close.

Canada faces a choice: step forward into the world, or be left behind.

There is a narrative about Canada that has been repeated so often it has started to feel like fact. That we cannot win. That our ambitions must go south to be taken seriously. That world-changing ideas need to be built somewhere else. History says otherwise.

There is a call for the brave, for the self-believing, and for those willing to swim against stagnancy. There is a chance to prove that Canadians can win.

In our roots, there is resilience. Canada was not built on excess. It was built on persistence.

There Is a Pattern in Our Roots

In the winter of 1921, a young surgeon named Frederick Banting walked into a University of Toronto lab with an idea that most of the medical establishment thought was wrong. He had no research budget, no proper lab space, and no track record in endocrinology. He had ten dogs, a borrowed room, and eight weeks before the facility needed it back. Despite everything, he discovered something that changed the world forever.

One year later, a 14-year-old boy named Leonard Thompson was dying of diabetes in Toronto General Hospital. His blood sugar was sky-high. He had weeks, maybe days. The clinical team administered Banting’s discovery, insulin, for the first time in human history. Within hours, Leonard’s blood glucose fell to normal. He lived for another 13 years, long enough to see the discovery that saved him earn a Nobel Prize.

That story is not an exception in Canada’s history. It is the pattern: a small team, limited resources, enormous ambition, and a result that echoed across generations.

Canada tends to narrate its own story with excessive modesty. We celebrate our politeness and our stability. These are real virtues. But we have grown apologetic, underselling our capacity for bold, world-changing invention. We have lost our own self-belief, and with it, our capability.

The truth is more interesting. Canada has been building at the frontier, in medicine, space, computing, and energy, for centuries. Not as a follower. As a pioneer.

Building at the Frontier

In 1950, while researching hypothermia treatments, electrical engineer John Hopps at the National Research Council made an accidental discovery: a stopped heart could be restarted mechanically or electrically. He built the world’s first cardiac pacemaker. It was large, external, and needed to be plugged into a wall, but the principle was correct. Today, over three million people worldwide carry a direct descendant of Hopps’ invention inside their chests. Canada gave them another heartbeat.

In 1975, NASA invited Canada to contribute to its Space Shuttle program. What Canada brought was a robotic arm, the Canadarm, a 15-metre articulated manipulator that became the most complex construction tool ever deployed. Built by Spar Aerospace (now MDA Space) in Brampton, Ontario, it flew on 90 Shuttle missions between 1981 and 2011. It helped assemble the International Space Station. It serviced the Hubble Space Telescope five times. When astronaut Scott Parazynski needed to repair a torn solar array in 2007 while standing at the end of an extended boom 90 feet above the ISS, it was Canadian technology holding him there.

Canadarm2, launched in 2001, has been a fixture on the ISS ever since. It has captured over 50 visiting spacecraft, including SpaceX’s Dragon and Northrop Grumman’s Cygnus. MDA Space is now deep into building Canadarm3, an AI-powered autonomous robotic system destined for Gateway, NASA’s lunar-orbiting outpost. Canada’s reach is getting longer. We are going to the Moon.

These are not stories of Canadian ingenuity waiting for American permission. They are stories of builders called upon to solve problems no one else could and who delivered.

The Long Revolution in AI

In the 1980s, when virtually every major research institution in the world had abandoned neural networks as a dead end, a British-born researcher named Geoffrey Hinton chose Toronto. He came because Canada was willing to fund speculative, long-horizon research through the Canadian Institute for Advanced Research. Work that the rest of the world had decided wasn’t worth the risk.

For decades, Hinton in Toronto, Yoshua Bengio in Montreal, and Richard Sutton in Edmonton worked in what felt like academic obscurity. They were studying the mathematics of how the brain might learn. Their colleagues thought they were chasing ghosts. The field was called “deep learning,” and virtually no one believed it would amount to anything practical.

In 2012, Hinton’s lab at the University of Toronto entered the ImageNet competition, the world’s most-watched test of machine vision. They won by a margin so large that it didn’t just beat the competition, it ended the competition. Neural networks weren’t a dead end. They were the beginning. The modern AI revolution, ChatGPT, Gemini, Claude, and every large language model you’ve heard of, runs on architectures descended from work done in Toronto, Edmonton, and Montreal.

In 2018, Hinton, Bengio, and Yann LeCun shared the Turing Award, the Nobel Prize of computer science, for their foundational contributions to deep learning.

Canada did not stumble into AI leadership. It invested in it patiently, for decades, when no one else would. Mila, the Quebec AI Institute co-founded by Bengio at Université de Montréal, is now the largest academic deep learning research centre in the world, with over 1,400 specialized researchers. The Vector Institute in Toronto, co-founded by Hinton, hosts over 1,000 researchers. Together with Amii in Edmonton, these three institutes form a national AI infrastructure that other countries are still trying to replicate.

This is what Canadian patience, ambition, and public investment can produce. Not a fast flip. A foundation that changed the world.

From Research to Runway

Canada’s innovation has not stayed in the lab. Shopify began in Ottawa in 2006 when Tobi Lütke, frustrated by the state of available e-commerce software, built his own. It now powers approximately 5 million stores globally, with a market capitalization of over $200 billion. Lütke, when asked why he hasn’t moved to the United States, replied simply: “Because I like Canada.” In a period when Canada’s sovereignty is under real external pressure, those four words became a rallying cry.

Cohere, founded in Toronto in 2019 by Aidan Gomez, Ivan Zhang, and Nick Frosst, has become the country’s flagship enterprise AI company. When a large US firm offered to acquire them at a nine-digit price, the founders refused. Today, Cohere is valued at over $7 billion, with revenue growing more than three times over in 2025. CEO Aidan Gomez was direct about why: “I really feel that our country is under threat. I feel that the box of Canada’s sovereignty has been opened, and it can’t be closed.” That is not the language of a startup pitch. That is the language of a builder who understands what is at stake.

Waabi, founded by Raquel Urtasun, who previously led Uber’s self-driving research, is taking a simulation-first approach to autonomous trucking, training its AI in virtual environments rather than running millions of potentially dangerous road miles. Backed by Uber and Khosla Ventures, Waabi began deploying fully driverless trucks in 2025.

Then there is Reaction Dynamics, a Montreal-based rocket propulsion startup building Canada’s capacity to reach orbit independently. A small team, few resources, a hard technical problem, and a mission tied directly to national sovereignty. Canadian Dynamism in practice.

The Unifying Thread

Banting worked on borrowed time in a borrowed lab. Hopps was researching something entirely different when he found the pacemaker. The Canadarm was Canada’s entry ticket to the Space Shuttle program, pragmatic, collaborative, and technically extraordinary. Hinton and Bengio funded deep learning research through Canadian institutions for decades before the world recognized what they had built.

Canadian innovation tends to be long-term. It tends to be patient. It tends to emerge from people willing to work on hard problems, in unglamorous conditions, for timelines that nobody else believes in. It does not tend to be loud. That is both its strength and its weakness.

The strength: Canadian builders build things that last. Insulin. The pacemaker. The Canadarm. Shopify. The foundation of modern AI.

The weakness: We have been better at invention than ownership. We discovered insulin and sold the patent for one dollar each. We conceived the light bulb and sold it to Thomas Edison. We built some of the most important technologies of the 20th century and watched the commercial value accrue abroad. The world benefited, but Canada did not always capture what it created. A sapling that dies before it has the chance to develop its own strength is lost potential. We forced our greatest inventions to move to survive.

Canada needs to start owning, protecting, supporting, and being proud of what makes it great.

The Moment We Are In

The external pressure Canada faces, economic, geopolitical, and technological, is real. But it is also an invitation. An invitation to stop outsourcing our ambition. To stop equating “not losing” with winning. To remember that we have grown world-changing inventions in the fields where no other crops could survive.

The resources were never lavish. Capital has been leaving quietly for two decades. In 2005, Canada’s largest pension fund held 74% of its investments at home. Today it holds 12%. The builders have noticed.

The next generation of builders in AI, in space, energy, defence, biotech, and climate, are not starting from scratch. They are standing on the shoulders of Canadian resilience. On Banting, Hopps, Hinton, and Bengio. They are inheriting a tradition of patient, rigorous, and frontier-facing work that has quietly shaped the modern world.

They need capital that leads, not follows. They need investors willing to sit with hard problems and long timelines. They need an ecosystem that treats ambition as a Canadian value, not an export. Building here is not a consolation prize. It is the point.

There is an opportunity to triumph. We have struggled long enough. Winter is over.

What Canadian Dynamism Means

American Dynamism is explicitly about the national interest, building for America, strengthening America, and making America more competitive. That is a worthy and coherent philosophy.

Canadian Dynamism is something different. It has always been oriented outward. Banting didn’t sell insulin to Canadians. He gave it to the world for a dollar. The Canadarm didn’t build a Canadian space station. It built humanity’s. Hinton and Bengio didn’t create AI for Canada’s benefit. They published their research openly, and the entire world built on top of it.

That is the Canadian instinct: to build things so good and so fundamental that the world cannot do without them. Not to win at the expense of others. To build in a way that makes everyone better off and to be indispensable because of it.

It means backing founders solving hard problems. It means recognizing that resilience of supply chains, infrastructure, sovereignty, and health systems is not a niche thesis. It is the defining challenge of the next decade. It means understanding that the country which built the tools that assembled the International Space Station, gave the world insulin, and quietly invented the foundation of modern AI, is not a country that needs to apologize for its ambitions.

It means building flagships. Not subsidiaries. Not branch offices. Not features inside external platforms. Companies of global scale, owned and led from Canada, solving problems that matter and keeping the value at home.

The Canadian Moat

Dynamism is resilience, a deep patience for difficult problems, combined with a certainty that through hard work, the problem will yield. Patience is the Canadian moat.

Banting didn’t give up when the university laughed at his proposal. Hinton didn’t give up when the field of neural networks was effectively declared dead. MDA didn’t stop building after the Canadarm. They built Canadarm2, then Dextre, and now Canadarm3.

That is the spirit. The one still standing when the problem finally yields.

Canada has always been a nation of builders. We built at the frontier of medicine when it was dangerous and speculative. We built at the frontier of space. We built the intellectual foundations of AI when the rest of the world thought it was a waste of time. We are now building autonomous systems for the Moon, enterprise AI companies with billion-dollar valuations, and rocket propulsion for sovereign access to orbit.

The pattern is clear. The capacity is real. The question is whether the next generation of builders will be supported, with capital, community, and conviction, to build at the scale Canada is capable of.

The future must be claimed. There are flagships to be built.

Ethan Marcoux is a writer and researcher focused on Canadian technology and defence innovation, and the founder of The Canadian Explorers publication.

Amiral Ventures is a Montreal-based seed and pre-Series A venture fund backing Canada’s most ambitious founders in AI-native enterprise software, resilience, and sustainability.

Welcome to the first edition of the Amiral newsletter.

We started Amiral to back founders who see what others miss and build where others won’t.

For our first issue, we’re starting with Patrick Murphy, founder of Maket. He’s making residential architecture accessible to anyone, no CAD, no $10K design retainer, just describe what you want. It’s a massive opportunity hiding inside one of the largest, most fragmented, and least digitized industries in the world.

Patrick is exactly the kind of founder we built Amiral for.

You’re receiving this because you’re part of the Amiral network. Not relevant? Unsubscribe here.

A warehouse on the outskirts of Montreal was the not the kind of place a father takes his 14-year old son.

Patrick Murphy sat in the passenger seat, watching the city’s polished neighbourhood dissolve into industrial grey, trying to keep his nerve. He asked for this trip himself, explained his idea with enough conviction that his father had listened, driven across town, and walked through a an industrial factory that looked like it was a hideout for criminal activity. Inside, among bolts of fabric and racks of blank shirts, Patrick picked out his production run while his father paid for it. The logo was already finished, as Patrick had built it himself, on Illustrator.

He called the brand Ambassador Apparel, because they would be ambassadors for Christ. Even at 14 he understood that a name should carry meaning and weight. He took the shirts to school and sold them in limited drops, manufacturing scarcity before he had a word for the strategy. He never lost money, but never really made any either. Each run broke more or less even.

This was Patrick’s first entrepreneurial experience, and although it was not a success story by any conventional measure, it proved something else. That he could originate an idea, build it into something physical, find the people who wanted it, and moved it from his hand to theirs.

Childhood

Otterburn Park sits on the South Shore of Montreal, a quiet city beside the hiking trails and rocky face of Mont-Saint-Hilaire, close enough to feel the mountains and far enough from downtown to breathe. Patrick was born there in 1994, the youngest of five children. In a large family, birth order shapes personality as surely as anything else does, and the youngest child often arrives already partially formed by the weight of the personalities stacked above them. In Patrick’s case, the defining characteristic of being last was that everyone left him alone.

“They always allowed me space to just be Patrick,” he said. “I was doing entrepreneurial stuff, I had a ton of ideas, and they were always really supportive.” He was not pushed toward medicine, law or any of the respectable professional tracks that families sometimes mistake for ambition. Whatever path opened in front of him, he would have to find it himself.

His father had grown up in California, studied theology in the United States, and then completed his PhD in Quebec, arriving in Francophone Canada in the 1980s to run a master’s-level training program for pastoral students. His father spoke no French. He was, Patrick would later say, about as locally intelligible as someone who had just landed in rural China. And yet he stayed, built a ministry, raised a family, and ran his life with the independence of a man who reports to no one. Patrick did not see the word “entrepreneur” written anywhere in his house growing up, but he was watching one in real time. His mother was an artist, moved by the Group of Seven deeply and permanently. She painted and created throughout Patrick’s childhood, and her aesthetic sensibility became something she passed on without trying.

Growing up in a house with five children on one salary and a mother working part-time as a teacher and artist, comfort was not the operative word. They never felt poor, however. The Lord always provided, as his father would say, and Patrick tends to agree. But there was not a lot of extra. There were five kids, one income, and a life in ministry that demands the soul in ways it does not always reward financially.

Patrick watched his father navigate this with the specific equanimity of someone who has made peace with uncertainty, and something about that lesson settled into him deeply. “The shared experience of an entrepreneur and a missionary is actually really close,” Patrick said. Both are trying to make something work with insufficient resources and an audience that isn’t always there. Both go through long stretches where nothing seems to be growing and where the natural response is to wonder whether they should have done something else entirely.

What Patrick discovered at 14, running Ambassador Apparel out of his school locker, was that he was a communicator and a salesman in the sense that he liked the work of persuading people, the social texture of it, the reading of what someone wanted and then offering it to them. He was not a scientist or a mathematician, and rather than spending years fighting against his nature, he leaned into it. He built the brand because he could build it. He sold the shirts because he was good at selling. He understood the mechanics of scarcity before he understood the term. And when each run broke even, he didn’t view this as failure. He viewed it as confirmation that the machinery worked, and that the next iteration would improve.

What also became clear, in those same years, was that his sense of design was not casual. He had built the Ambassador Apparel logo himself because it had not occurred to him to outsource it, and because he had opinions about fonts, spacing, and a hunch on what a good logo communicated versus what a sloppy one did. He was the kind of kid who cared about what things looked like, genuinely cared, the way his mother cared about her paintings.

University

After high school, rather than enrolling in a traditional university program, Patrick went to CEGEP for graphic design. He wanted the technical vocabulary of design so that he could do contracts, generate income, and develop skills that would make him useful rather than dependent. He was not interested in a degree for its own sake. He wanted tools.

The three-year program gave him those tools, and after CEGEP, he went to Concordia University for communications, specializing in sound production, rounding out a creative education that was broader than most people would have chosen.

During those years, while completing coursework, he was also working full-time at a performance marketing agency. He had come through an unexpected door, where Patrick had started a hockey blog called Hockey Busts, covering draft prospects with earnest obsessiveness. The agency needed someone to run a hockey blog for their clients. As matches go, it was almost embarrassingly convenient.

He joined as the one of the first employees. He was in his early twenties, recently out of design school, and knew very little about digital marketing. His boss, James, knew this and hired him anyway. Patrick would later describe this as the defining professional gift of his early career, a boss who extended responsibility before it had been earned, who let mistakes happen, who trusted Patrick with increasingly important pieces of the business and then stepped back to watch what happened.

“James just gave me a shot,” Patrick said. “I was just coming out of university. I did not know anything about anything when it came to digital marketing. And he allowed me to run a very important part of his business.” The lesson Patrick absorbed was less about marketing specifically and more about what good leadership looks like from the receiving end. A mentor who extends trust, even when the trust is not yet fully warranted, creates a different kind of professional than one who makes people earn every inch of responsibility.

At the agency, Patrick learned the mechanics of performance marketing in ways that would serve him a decade later, on how to drive inexpensive traffic, how to measure conversion, and how to think about customer acquisition costs in relation to customer value. He worked across clients in different industries, including a shoe company obsessing over return on ad spend, a solar company needing high-quality B2B leads, a fintech application fixated on cost per install. Each client was a case study in what a different kind of business needed to grow, and Patrick was taking notes on all of it. He also became certain that he did not want to be in the services business forever, and that he wanted to build something of his own.

COVID

The transition came through ThirdBridge, a software development shop with ambitions to build an office management platform. Patrick joined as head of commercialization, a role that suited his skills and gave him real exposure to the work of building software. He had ideas about products, always had, but he did not know what it cost to build them, how long it took, what testing looked like, what deployment meant, and how things broke when real customers got their hands on them. ThirdBridge gave him that education.

He worked closely with the company’s CEO, Pierre-Étienne Bousquet, who trusted Patrick with access to leadership conversations that a typical employee at his level would not have seen. Patrick was thinking alongside the founder, not just executing for him. He learned what it meant to make a product decision, commit to a technical direction, and discover mid-build that the thing you were building wasn’t quite right.

ThirdBridge’s platform was aimed at a problem that COVID rendered immediately irrelevant, the friction of office check-in solutions, conference room booking, and shared physical space. When the pandemic hit, the company cut the initiative. Patrick found himself, for the first time, genuinely unemployed and free.

There was someone he had been thinking about for a while. Years earlier, back when he was still at the performance marketing agency, a man named Stéphane had reached out to him. Stéphane had grown up in Montreal, moved in overlapping circles with Patrick without the two of them ever becoming close, and been running his own architectural design agency.

Stéphane had an idea for a SaaS platform and was looking for a co-founder. Patrick told him no, or something close to no. The timing was wrong, and Patrick needed more experience, learn more about software, and become more useful before he could be useful to someone else. They stayed loosely in touch. After ThirdBridge went quiet, Patrick called Stéphane.

A month after he had stopped going into an office, ThirdBridge called Patrick back. They wanted him to return. He sat down with his wife, who had been with him since university and who understood, what kind of person she had married, and told her what he was thinking. He wanted to build a company, but he wasn’t sure it would work and if he was ready. She listened and then asked, “What’s the worst that happens? You try this and it doesn’t work, you can go do something else. And the best that happens is that you try this and it works and you’re really happy.”

He describes her as having kicked him off the ledge. He was standing at the edge of something, hesitating, and she pushed. He turned ThirdBridge down and called Stéphane back.

Maket.ai

They started the company during COVID, working remotely from the beginning. Neither of them was an engineer. They were two business-oriented founders with ideas, relationships and the ability to sell, but without someone on the team who could build the underlying technology. They were, in fact, constructing a pitch without the product to back it up. They got into NEXT AI, one of Quebec’s most prestigious startup accelerators, on the strength of a personal relationship Patrick had with the program director. By their own admission, they did not belong. “We had no idea what we were doing,” Patrick said. “We’re two founders. We’re not technical. You had to have a technical founder and a business founder to get in. We’re both business guys.” They got in anyway. The program introduced them to someone at Mattamy Homes, Canada’s largest home builder.

Mattamy had been thinking internally on whether AI could generate floor plans.

The question landed at a moment when Patrick and Stéphane were open to a new direction. Patrick looked around at the housing stock in Otterburn and told Stéphane, “Dude, everything is ugly. I hate these houses, they look terrible. Everybody’s living in terrible garbage boxes that all look the same. Why does this space around us not look nice?”

Mattamy’s suggestion aligned almost perfectly with that irritation. And yet the pivot was not easy. The technical challenge of using AI to generate floor plans was, as Patrick and Stéphane would quickly discover, much harder than it appeared. “Nobody’s doing this,” they had thought when Mattamy posed the question. “It must not be that hard.” It was, in fact, incredibly hard.

The argument over whether to make the pivot happened on a phone call while Stéphane was hiking up a mountain, the signal cutting in and out, both founders raising their voices across a dropping connection. Stéphane wanted to stay the course on their original thesis of an online home sales platform. Patrick wanted to go all in on generative architectural design, on the harder, stranger, more exciting path. You do not know who your business partner really is until you have disagreed with them about something that actually matters. They resolved it by following the money, or the prospect of it. Mattamy was willing to be their first paid pilot. A real first customer changes the economics of almost any argument. They took the generative design path, secured the contract, and used that traction to get into Techstars.

Distribution

What Patrick understood before Maket had a good product was that good products need customers to become good, and customers do not appear on their own. He had been watching this mechanic for years, at the agency, at ThirdBridge, in his own early ventures, and he had come to believe that distribution was not a problem you solved after the product was right. Distribution was something you built alongside the product, sometimes ahead of it, because without the signal that real users generate, you are building in the dark.

Rather than following the conventional early-stage advice of cold calls, warm introductions, and careful qualification of prospects, Patrick and Stéphane spent money on Facebook ads. They set up landing pages, tested messaging, drove traffic to a waitlist, and watched what happened. What happened was that people signed up. A lot of people. By the time the V1 of the product was mature enough to support real users, Maket was seeing roughly 1,000 sign-ups per day.

The aggregate spend over the course of a year came in part by a commercialization subsidy from the Quebec government that Patrick deployed. “The government is giving us free money,” he reasoned, “we’ll spend it on ads.” The marketing drove organic search, which drove more sign-ups, which made attribution complex, but it also generated data that would be more valuable than the numbers. Feedback forms, customer support logs, live chat transcripts, and churn exit surveys were the holy grail of the government money spent.

Olivia, who joined Maket before she had finished her university degree and would still be with the company three years later as she completed her final semester, managed the customer-facing edge of this feedback loop with the specific thick-skinned patience of someone who has absorbed a great deal of complaint on behalf of a product that was not yet what it needed to be. She knew the customer the way someone knows a difficult family member, deeply and sometimes painfully. What came out of all those complaints, feedback forms and support emails was a portrait of what the platform needed to become, a portrait that would eventually shape the architecture of V2 in ways that a hundred curated customer interviews could not have produced.

One of the key lessons Patrick absorbed at the performance marketing agency was less about marketing specifically and more about what good leadership looks like from the receiving end. A mentor like James who extends trust, even when the trust is not yet fully warranted, creates a different kind of professional than one who makes people earn every inch of responsibility. Patrick, now in the mentor seat, would extend that same trust to Olivia, building a knowledge base of the customer that made her indispensable.

Fundraising

The fundraising process lasted 14 months, produced 100+ rejections, and came within weeks of ending the company.

Patrick began preparing for the seed round in Q3 of 2024, at a point when Maket was on an upward trajectory but carrying a structural vulnerability that any experienced investor would identify immediately, that the product was not good. The V1 needed a lot of work, but the initial concept was presentable and functional. The team was two business-oriented co-founders without a strong technical leader, which meant that Maket had distribution and market validation but could not fully capitalize on either. Patrick understood this, but decided to run a process anyway for survival.

He compressed the meetings into the shortest window he could manage, stacking them back to back in August and stretching through November, creating the artificial urgency that a tight timeline produces in any serious negotiation. Inbound came from tier-1 investors. Conversations went deep, some progressing all the way to investment committee presentations. And then came the rejections, one after another, patient and varied in their specific wording but consistent in their underlying message: a product that wasn’t good enough and a founding team without a technical founder.

In July of 2024, a message arrived from someone Patrick had not been expecting to hear from. The man was a successful entrepreneur who had sold his company and was now looking to deploy capital into things he understood. He was interested in Maket, and they talked several times over the following weeks. The conversations went well, which is to say they went slowly, because the man seemed to move at his own pace and respond on his own schedule.

In August, Patrick was on a flight to Japan. He was taking a real break, the kind that founders rarely allow themselves and that Patrick had learned, partly through experience and watching his father manage exhaustion through decades of ministry, to actually take. The message came through during the flight and the investor was in. He would be the lead, the round would be anchored, and the long process would be over.

Patrick landed in Japan, spent the rest of the trip doing data room and term sheet work between stretches of rest. When he flew back to Montreal, he told his team they were good to go.

And then the man stopped responding.

Patrick followed up. He reached out to mutual contacts, people who knew the investor and had vouched for him. They described him as a great investor, someone who would be terrific on a board. He just wasn’t responding. A few weeks passed, then a month. It became clear, slowly and then all at once, in the way that a funding round falls apart, that the yes had been a no that hadn’t yet been delivered. Maket had been operating on the assumption that the round was closing while the round was, in fact, dissolving.

“We were about to shut down the company,” Patrick said, “because we almost ran up our runway.”

He met with Amiral in January of 2025. Patrick had first encountered Nectar about 2.5 years earlier at a mixer in Montreal’s startup ecosystem, when the fund was still in the process of being raised. He had stayed loosely in touch. Later, Fred had become a kind of informal mentor through a Quebec Tech program. Patrick met Fred for coffee and told him the truth, all of it, including the ghost investor, the near-shutdown, and what had gone wrong. At one point, Patrick told Fred directly not to invest. “I’m talking to you as someone I consider a close friend. I don’t think you guys should invest.”

Fred never forgot that. The honesty of it, the willingness to prioritize a relationship over an opportunity that might have closed a desperate gap, was the kind of signal that stood out as something durable. When Amiral was capitalized and the timing was right, that meeting was still sitting in the ledger of things Fred knew about Patrick.

The round eventually came together. Blitzscaling Ventures closed on their commitment in January of 2025, while Patrick was back in Japan, but this time closing their now largest customer. Their partners applied what Patrick described as a formula rather than a gut feeling, that the company had distribution, the ability to blitz a market, and disrupt technology within the space. Jeff, the partner who championed the deal from Blitzscaling’s side, had a single quality that Patrick would describe as the thing that mattered most after hundreds of rejections. He asked thoughtful questions and understood what he was hearing. The conversation that closed the round was, Patrick said, “very, very easy.” Amiral Ventures ultimately led the $3.7 million seed round alongside Blitzscaling, Hidden Layers, BYVP, StartUp in Residence, Spatial Capital, and others.

Bruno Morency, an advisor and investor through Techstars, had been with Patrick through much of the navigation, helping him think through the negotiations and the moments of uncertainty. “Bruno really, really helped me a lot during fundraising,” Patrick said. The visible part of fundraising is the pitch; the less visible part is the person in the corner helping you figure out what to do next, and Patrick is someone who is honest enough about what he doesn’t know to go find those people.

“It’s like a long sales cycle,” he said, “and you’re knocking on hundreds of doors and you just have to knock on one and find somebody who has conviction. It’s all like dating, right? You meet somebody for 15-30 minutes, and you know if they’re leaning in.”

Ski Crash

Three years before the seed round closed, in 2022, Patrick crashed on a ski slope and fractured his femur badly enough that paralysis seemed possible. He recovered. The bone healed, and continues to ski on the weekends.

But something recalibrated in the accident’s aftermath, in the weeks of recovery and the months that followed. Patrick had started Maket in 2020 and had been building it for two years when it happened. He was the same person after the crash that he had been before it, curious, optimistic and energized by design and by the problem he was trying to solve. But the crash gave him a counterweight to the ambition, a second truth to hold alongside the first. “Every day is like a plus-one bonus for me,” he said.

What he means by this, in the context of building a company, is nuanced enough to be worth unpacking. He is not saying that the accident made him less ambitious, or that the brush with paralysis made him indifferent to outcomes. He is saying that it forced him to hold two truths simultaneously, the truth that what he is building matters and demands everything he has, and the truth that no amount of what he has can guarantee the outcome, because the fragility of life does not negotiate with ambition. “No matter how hard you build,” he said, “there’s a certain element of just luck and right place, right time and blessing, whatever you want to call it, that determines some of the paths.”

His father’s advice during the dark periods of building Maket would draw directly from decades of ministry experience, to give yourself three-month marks. Set a start time, check in at three months, see where you are. If you still feel like quitting, allow the feeling but don’t act on it. Then do it again. “It would be the same thing of when he was in ministry,” Patrick said. “He would just always take a reassessment.” The advice sounds simple but it is structurally very good, because it takes the chaos of a difficult stretch and turns it into a data collection exercise. You are not deciding whether to quit; you are deciding whether the evidence at a specific checkpoint supports continuing. That reframe saved Patrick more than once.

Fun as a Moat

Patrick has been building Maket since 2020, which is a long time to be doing something that started, by his own admission, as a guess. The guess was that AI could do something for architecture, interior design and floor plan generation that had never been done before, that there was a market of homeowners, builders, real estate developers and agents who were underserved by existing tools, and that the combination of distribution and technology could unlock something large. Five years in, with a V2 about to launch, a seed round closed, 30,000 sign-ups per month, and 50,000+ users waiting for the next version of a product they were already using, the guess looks increasingly like a thesis.

His competitive advantage, as he sees it, is that he is having fun. He quoted a tweet to that effect, couldn’t remember exactly who wrote it, but felt it described him accurately enough to repeat. As competitors have multiplied in the generative design space, the fun has acquired a sharper edge. He is nice, he said, but also very competitive, and when he sees a competitor doing something, his instinct is to figure out how to counter it. The fun and the competitiveness are not in tension; they are the same thing, because he is playing a game he has chosen and cares about, rather than one he has been assigned.

He has resisted the pull of San Francisco, not without understanding what it offers, but with a specific argument against it that a mentor once made to him, that building in Montreal, with less access to capital and therefore a slower ramp, meant that when the technology matured enough to match the ambition, the company had not yet spent all its money chasing a vision the market wasn’t ready for. The slower ramp had been, in retrospect, a form of protection.

He believes Montreal is due for a rise on the global stage, that the concentration of AI talent and the quality of the institutions feeding it create the conditions for something significant.

His father lives nearby, and discovered recently, that AI can do extraordinary things. He uses Claude to write theology books, gets them approved through an e-publisher, creates songs with Suno, generates images. Patrick watches this with the specific delight of someone who has spent 6 years trying to build something what his father is now using, and who sees, in his father’s creativity finally flowing freely through the right tools, a version of what he is trying to give to everyone. “People who thought that they weren’t creative, now they can be creative.”

Maket is building the platform for generative architectural design. If you’re a homeowner, home builder, real estate developer, or agent who needs to create and modify floor plans, check out Maket.ai.

Share

Designing a home has always been a process of vision meeting constraints, a dance between creativity, regulations, and technical feasibility. Maket is reimagining that process.

Founded in Montréal by Patrick Murphy, Stéphane Turbide, and Simon Vallée, Maket empowers anyone, from architects and builders to homeowners, to generate and refine architectural floor plans in seconds. Users can describe what they want in plain language (“a three-bedroom house with an open kitchen and lots of natural light”), and Maket instantly generates multiple, code-aware design options ready to explore in 2D or 3D.

What Canva did for graphic design, Maket is doing for architecture: democratizing access to creation.

From Vision to Platform

The idea began with a simple question: What if designing a home could be as easy as sketching a thought?

Patrick and Stéphane, both passionate about architecture and user experience, started developing the core generative engine in partnership with Mila, Montréal’s world-leading AI institute. Patrick brings a background in design and growth, shaping Maket’s path from product to market. Stéphane spent over a decade as an architectural technologist, witnessing firsthand the cost and complexity faced by non-architects. Their third co-founder, Simon Vallée, a veteran product leader and three-time founder with exits to Groupon (OpenCal), Slack (Spaces) and Figma (Flight), joined to help shape that vision into a world-class product.

Today, Maket’s platform enables users to instantly create, edit, and visualize realistic floor plans. Behind the simplicity lies a complex generative AI system that learns from every project created, improving with each iteration. With over a million users and millions of plans generated, Maket is building a powerful data flywheel and a deep technical moat.

The Next Chapter: From Consumers to Builders

As Maket grows, its mission of democratizing architectural design extends to professionals, empowering builders and developers with the same intuitive AI tools that made design accessible to everyone.

Its proprietary models, refined through millions of user interactions, now serve as the foundation for a professional-grade platform that can compress weeks of front-end design into minutes. Builders can generate compliant plans, adapt them to local codes, and instantly visualize options for clients all while reducing costs and accelerating time-to-market.

This evolution positions Maket not only as a creative tool but as a core design infrastructure for the construction industry, bridging the gap between design intent and build execution.

AI Meets Architecture

Maket’s core innovation lies in its AI-powered design engine, which combines deep learning, architectural rules, and reinforcement learning to generate realistic, buildable layouts. The system understands adjacency constraints (like placing bedrooms near bathrooms), respects local zoning and building codes, and can even suggest style variations on demand.

By compressing weeks of early-stage design work into minutes, Maket delivers tremendous efficiency for professionals and creative freedom for individuals. For architects and home builders, it’s not just a productivity tool; it’s a new way to think about design.

A Team Building for the Long Term

Maket’s founding team brings a rare blend of expertise:

• Patrick Murphy (CEO) blends creativity, product vision, and storytelling.

• Stéphane Turbide (COO) grounds the company in deep architectural and technical rigor.

• Simon Vallée (CPO) brings two decades of experience building intuitive and category-defining products with global reach..

Together, they’re shaping an entirely new category at the intersection of architecture, design, and AI.

Why We Invested

At Amiral Ventures, we back founders using AI to reinvent traditional industries. Maket represents that vision perfectly: a category-defining company born in Québec, tackling one of the world’s most entrenched, analog sectors, architecture and construction.

We believe the next generation of design software will be conversational, data-driven, and generative. Maket is leading that shift. The company’s technology, user growth, and ambition position it to become the global reference point for AI-assisted design, a new standard for how spaces are imagined and built.

CFOs at every mid-market and enterprise company are constantly seeking to better understand the operational and financial performance of their business; which product SKUs are less profitable? Why is my inventory cost higher in certain areas? Why am I paying more overtime in specific teams? The answers to these questions are well hidden within their multiple Enterprise Resource Platforms (ERPs) and other financial and operational systems of record: NetSuite, Oracle, Salesforce, ServiceNow, Quickbooks, SAP... There are thousands of those systems and it is common for an enterprise to use several if not dozens of them.

In order to answer these daily questions, CFOs and FP&A professionals end up spending their days navigating the complexity of multiple systems, extracting complex, disjointed data from several disparate ERPs, and attempting to join them in Excel. This is time consuming, error prone, and incredibly inefficient. ERP providers’ BI and analytics solutions are siloed and proprietary, and don’t move at the speed of business, simply put: it sucks.

Maxa was founded in 2019 to help solve this problem hiding in plain sight. Maxa enables the complete automation of a company’s ERP data, across all systems, so that FP&A teams have a single source of truth and can deliver insights in a snap.

Building a magical solution

Maxa automatically unifies multiple ERP and data systems into a data product. A Maxa Data Product is enterprise data that is standardized, made relevant with pre-computed insights ready to be consumed by non-technical users and reusable across tools and algorithms, without the business logic being locked into a visualization layer. All these insights are generated in seconds, not weeks.

It’s all delivered via the Snowflake Data Cloud as a Snowflake Native App. Maxa created the universal enterprise data schema which is essentially an intermediate mathematical language; they built a single data model across ERPs. This allows them to take siloed systems and make their data available in an efficient and unique way to advanced analytics and artificial intelligence models.

Founders on a mission​

Raphael Steinman and Alexis C. Steinman are the quintessential founders. Smart, gritty, humble, not afraid to take risks and looking to make a big dent in the world. They just raised a US$21M Series A in which Amiral Ventures participated - we couldn’t be happier. The company is on an incredible trajectory and is poised to become a Canadian flagship.

Why now?

The underpinning of Maxa’s solution depends on storage and compute becoming extremely cheap and infinitely scalable which only really became a reality in the past few years. Their solution hinges on the ability to crunch billions of rows of data in seconds. This allows for Maxa to create massive value for their customers. Their algorithm also becomes highly reusable. Hundreds of templates now become available for all their customers and insights available at the click of a button. In essence, they are trying to make advanced financial analytics a commodity; taking something that used to be really hard and allowing their customers to move much faster and run their business much better.

A scalable and robust model

Success comes from making good decisions (and learning from bad ones). One of the good decisions they made was to not try to reinvent the wheel. Instead of building a reporting front-end, Maxa simply leverages existing tools like Power BI, Tableau and Looker. In the current age of LLMs, this was quite prescient. One of the new opportunities for Maxa is to use this new form of computer communication instead of legacy dashboard reporting. The promise of generative AI only works if the data can be trusted and the single source of truth for companies is their ERP.

Moreover, it’s a usage based business model which is super scalable. Customers get hooked since it solves a major problem and decide to use Maxa for more & more analytics needs (read: strong LTV). Ultimately, this gets people out of crunching data in Excel and frees up executives’ time to work on the business. Demand planning, forecasting, procurement, inventory management all generate tons of data; what’s missing is insights. This is where Maxa excels. For example, live performance per SKU was not imaginable a few years ago. This drives automated insights which can be tied directly to financial impact for their customers. To top it off, their native app captures Snowflake’s capabilities all while protecting sensitive customer data.

The global opportunity

The ERP software market is substantial and rapidly expanding. An estimated $250B is spent on systems each year. it’s projected to exhibit a compound annual growth rate exceeding 24% over the next several years. This growth is fueled by the escalating need for operational efficiency, transparency in business processes, and the urgent requirement for data-driven decision-making. There are 4M companies globally running an ERP. Additionally, there is an estimated $1T staff automation cost savings potential with companies increasingly seeking manual task automation. It’s a limitless opportunity for Maxa.

Building the next Canadian Flagship

Maxa demonstrates exceptional potential for growth in a massive global market. Their innovative solution addresses a clear pain point which demonstrably saves an FP&A team significant time while demonstrating rapid return on investment for their customers. This creates a lot of value to their users with a robust and scalable business model. Maxa was also selected as startup of the year by Snowflake as well winning the 2024 Data Driver’s award which has engendered significant customer interest. The founding team possesses both deep industry expertise and a track record of execution, bolstering our confidence in their ability to drive Maxa’s success. This combination of product-market fit, scalability, and a strong team positions the startup exceptionally well for rapid growth.​

Maxa fits squarely into Amiral’s vertical AI thesis: using innovative data-first technology to create significant customer value. There is a clear imperative for businesses across the world to increase their productivity and we believe that tackling ERP data to derive real-time financial insights is a huge unlock to allow each business to focus on what really matters. It is clear that running LLMs on financial ERP data has huge potential if you can trust the results, and Maxa has demonstrated that running AI on the Maxa universal model provides better results than directly on the ERP’s complex databases.

Congratulations to Alexis, Raphael and the entire Maxa team. You are only getting started!

In this candid episode, the three partners trace the origins of the fund: how Fred's entrepreneurial journey (including co-developing LTE at Nortel and a successful Series A & B raise at Mnubo) first brought him together with Dom, a veteran M&A banker who went on to launch Xpnd Capital and deliver two IPOs. Nectar rounds out the team, coming off years of post-investment value creation work at PNR and crossing paths with Fred through this very podcast.

The conversation gets real about what it actually takes to raise a fund from scratch: nearly three years in the trenches, no product to show, just a slide deck and conviction. They share what they've learned, the primacy of trust-building, the importance of giving candid feedback to founders (even on a "no"), and the gap they're filling in the Canadian ecosystem at the seed-to-Series A inflection point.

The team also lifts the curtain on their Prosperity Decoded thesis, backing AI-native companies driving enterprise productivity, sustainability, and resilience; and spotlights their first two portfolio companies: Maxa (AI for CFOs and FP&A) and Maket (AI for residential architecture and design).

They close on what excites them most: the caliber of Quebec founders, the community of successful exits now back as LPs and mentors, and the long game of building Canada's next generation of flagship technology companies.

We talk about his journey from building robots as a kid to launching startups while still in university, what he learned from failing his first company, and how General Magic is rethinking the way humans interact with software using natural language.

We also dive into their recent $7M seed round, going through a16z Speedrun, and why the future of computing might move beyond the traditional graphical interface.

I really enjoyed this conversation, Anthony is thoughtful, ambitious, and building something genuinely interesting. Let's get into it.

Tyler shares his journey from space-obsessed kid to Stanford researcher to venture-backed space founder, and explains how Zona is building a next-generation positioning system designed for centimeter-level accuracy, stronger signals, and built-in security. We unpack how modern infrastructure depends on timing and location, why GPS is vulnerable to jamming and spoofing, and what new satellite constellations unlock for autonomous systems, defense, data centers, and everyday devices.

We also go deep on building a hard-tech startup: raising capital for space infrastructure, launching satellites, assembling the right team, and navigating the long road from research to orbit.

If you're interested in space tech, deep tech startups, autonomous systems, or the new space economy, this episode is for you.

Topics covered:

• Why GPS needs an upgrade

• How next-gen satellite navigation works

• Centimeter-level positioning and resilient timing

• Space infrastructure and low-Earth orbit constellations

• From PhD research to space startup

• Lessons from building a hardware + space company

• The coming wave of space innovation and lunar missions

Topics covered include: Early exit offers vs long-term ambition Scaling from startup to global platform Fundraising strategy and cap table discipline Hiring executives and leadership evolution Crisis moments and near failures Building and keeping tech champions in Canada AI, moats, and the future of company building Recorded live at North Star — Inspiring Entrepreneurs in Montréal.

Across all three segments, one theme dominated: AI is moving faster than any tech cycle before it, and Canada has a once-in-a-generation chance to lead.

VC Panel, Navigating the AI Wave

Investors from Inovia, White Star, and Amiral Ventures explored the chaos and opportunity of today's market. The AI cycle is "psychotic" in speed and scope, but the fundamentals still matter: strong teams, sticky products, real customer value. The best startups will use AI as an enabler, not the product, and differentiate through privileged data, deep domain expertise, and the ability to deliver ROI, not hype.

Founder Panel, Building Real AI Companies in Montréal

Budpress, Maxa, and Wrk shared candid takes from the front lines: why traditional SaaS is being rewritten, how open-source models and reasoning engines are changing the game, and why Montréal remains one of the best bases in the world for AI talent and cost-efficient scaling. The founders stressed that Canada must shed its "small market" mindset, big outcomes are possible here.

Keynote with Sam Ramadori, AI, Climate, and Responsibility

Sam recounted his unlikely journey from private equity to leading BrainBox AI through an acquisition, and why he's now dedicating himself to Law Zero, a nonprofit effort led by Yoshua Bengio to build safer, more reliable AI systems. His message was clear: Canada has extraordinary AI talent, but must rally around sovereignty, responsible innovation, and deeper collaboration to avoid being squeezed between global superpowers.

The night ended with a message to the community: this is our moment. We have world-class research, world-class founders, and a fast-maturing ecosystem. If we choose ambition, and support one another, Montréal can be one of the defining AI hubs of this decade.

Étienne opened up about the lessons he learned along the way, the value of resilience, trust, and playing the long game in both building and backing startups. He also spoke about his partnership with Luis & Varun at Telegraph Hill Capital and their shared mission to support founders in Quebec.

Real talk, honest stories, and practical takeaways, this one's for anyone thinking about their next chapter, whether it's starting up or stepping into VC.