Founder Story: Scott Wu of Cognition
The Prodigy from Baton Rouge
Scott Wu was born in 1997 in Louisiana to Chinese immigrant parents and grew up in Baton Rouge, where he attended Baton Rouge Magnet High School. He began coding at age 12, and the machinery of competition math and computer science took hold early. In 2011 he became the Mathcounts national individual champion, the country's premier middle-school math title, representing Louisiana against the best young mathematicians in the United States.
Programming quickly became his real proving ground. Wu won three consecutive gold medals at the International Olympiad in Informatics (IOI) in 2012, 2013, and 2014, taking first place overall in 2014. He was not competing in a vacuum: his older brother, Neal Wu, was himself a three-time IOI gold medalist, and the two grew up trading algorithms across the dinner table. On the Codeforces competitive-programming ladder, Scott reached the rarefied rank of Legendary Grandmaster with a peak rating around 3350. Few humans on earth have that combination of raw problem-solving speed and depth, and it would later become the DNA of the company he built.
Harvard, and the Decision to Leave the Track
Wu enrolled at Harvard University, where he kept competing at the highest level. As part of Harvard's team at the 2016 International Collegiate Programming Contest (ICPC), he won a gold medal and placed third overall in the world — the collegiate equivalent of an Olympic podium. He graduated in 2019, notably with a degree in economics rather than computer science, a small hint that Wu was already thinking about systems, markets, and incentives, not just algorithms.
Before Cognition, Wu did the unglamorous work of learning how companies actually get built. He was a software engineer at the fintech firm Addepar, then co-founded and served as CTO of Lunchclub from roughly 2017 to 2022, an AI-powered networking startup backed by Lightspeed, Coatue, and a16z. At Lunchclub he built the matching algorithms that paired thousands of professionals for one-on-one meetings. It was a real business and a real education, and it earned him a spot on the 2020 Forbes 30 Under 30 list. But it was not the thing. Wu's competitive instincts were pointed at a much larger target, and by late 2022 the ground beneath software was shifting.
Founding Cognition
In November 2023, Wu co-founded Cognition alongside Steven Hao (CTO) and Walden Yan (CPO). The pedigree was almost absurd: Hao was an MIT computer science graduate, a 2014 IOI gold medalist, and an early top engineer at Scale AI; Yan was a Harvard student who dropped out in 2023, had worked on Cursor at Anysphere, and won IOI gold in 2020. Collectively, the founding team held roughly 10 IOI gold medals — a fact that became part of Cognition's identity and its recruiting pitch. If any group of people could teach a machine to reason about code, this was a plausible candidate.
The path was not linear. The team initially explored cryptocurrency projects before pivoting to generative AI in late 2022, chasing the sudden capability jump in large language models. The founders bet that the frontier was not another chatbot but an agent — a system that could plan, execute, debug, and iterate on real engineering tasks end to end. That thesis attracted early conviction capital: Peter Thiel's Founders Fund led a roughly $21 million round, reportedly at around a $350 million valuation, before the product was even public.
The founding team held roughly 10 IOI gold medals — Cognition was less a startup than a collection of the world's best competitive programmers pointed at a single question: could you build the engineer, not just the tool?
Building Devin
According to accounts of the early days, a defining moment came just before Christmas 2023, when an early version of Devin autonomously resolved a server configuration issue the team had been stuck on. That was the signal that the agent could do more than autocomplete — it could own a problem. On March 12, 2024, Cognition introduced Devin publicly as "the first AI software engineer." The launch demo detonated across the internet, racking up more than 30 million views on X and triggering an industry-wide debate about the future of programming almost overnight.
The claim had numbers behind it. On the SWE-bench benchmark of real GitHub issues, Devin resolved 13.9% of tasks end to end, against roughly 1.7% for GPT-4 and 4.8% for Claude 2 in the same evaluation at the time. Unlike a coding assistant that suggests the next line, Devin was designed to work like a colleague: read a ticket, plan an approach, write and run code, test it, and fix its own mistakes. Wu has been careful to describe the machine's competence honestly, placing it "somewhere between a junior and a mid-level engineer" depending on the task. Where it shines is not glamorous greenfield invention but the toil that burns out human teams:
- Long-tail maintenance and bug-fixing
- Updating and modernizing legacy software
- Platform and framework migrations
- Repetitive, well-scoped tickets that never end
The Augment-Not-Replace Philosophy
Wu's most consistent public message is that Devin is a partner, not a pink slip. As AI-coding fears reached a fever pitch in 2026, he pushed back directly:
"We've never thought about it as replacing humans. I know it's like a scenario, folks have said these things. It has never been our view."
His framing is almost tender for a benchmark-obsessed competitor:
"When we started building Devin, it's kind of a funny thing, but we really just thought of it as: this is your buddy who helps you build more."
That philosophy shows up in his design principles. Wu insists that "it should always be up to the human what to do," keeping engineers in the driver's seat while the agent handles the grind. He grounds the whole project in why people write software in the first place. "Most software engineers love building software," he has said. "If you ask them why, what they'll basically tell you is, 'Well, it's like I get to build things from nothing.'" In Wu's telling, offloading tedious maintenance to Devin is what frees engineers to keep doing the part they love — the creative, architectural, from-nothing work.
Rocketship Growth and a $26B Valuation
The market rewarded the vision at a pace few companies have matched. By September 2025, Cognition had raised a Series C led by Founders Fund at a $10.2 billion valuation. Then, in May 2026, the company announced it had raised more than $1 billion at a $25 billion pre-money / $26 billion post-money valuation, in a round led by Lux Capital, General Catalyst, and 8VC, with participation from Elad Gil, Founders Fund, Ribbit Capital, and others. In roughly eight months, its valuation had more than doubled.
The revenue told the same story. Cognition's annualized revenue run-rate went from about $37 million in May 2025 to roughly $492 million in May 2026 — a more than 13-fold jump in a single year — powered by enterprise usage growing about 50% month-over-month for six straight months. Along the way, Cognition acquired the remaining assets of Windsurf in 2025, absorbing a leading AI-coding IDE and team. Its customer roster came to include names like Goldman Sachs, Mercedes-Benz, and multiple U.S. government agencies. Perhaps the most striking proof point was internal: by 2026, the large majority of the code committed at Cognition was reportedly written by Devin itself — the company had become its own most demanding customer.
Legacy, Skeptics, and What Comes Next
Not everyone is convinced. Devin's early demos drew scrutiny over how much was truly autonomous, and skeptics have questioned whether a $26 billion valuation can be justified in a market crowded with well-funded coding-agent rivals. Wu's answer has been to keep shipping and to keep the framing human. He has predicted that agents will spread into customer service, medicine, and beyond, but returns each time to the same guardrail: they should augment, not displace.
What makes Wu a distinctive founder is the through-line from the IOI to the enterprise. He treats software the way a competitive programmer treats a hard problem — decompose it, automate the mechanical parts, and reserve human ingenuity for the genuinely novel. He has described a coming "Moneyball-ification" of engineering, where measuring and multiplying what each person can build reshapes how teams are structured. If he is right, the legacy of Cognition will not be that it replaced engineers, but that it gave every engineer an infinite intern that never sleeps — and forced the entire industry to rethink what a software team is for.
Closing Thoughts
Scott Wu's arc is a study in refusing the obvious next step. He left the competitive circuit at its peak, left Harvard's prestige track, and left a perfectly good first startup — each time because a bigger, harder problem was pulling at him. The competitive-programming childhood was not a quirky biographical footnote; it was the training ground for a company built to reason about code at superhuman speed, staffed by a founding team with an almost comical concentration of world-champion talent.
Whether Cognition ends up worth $26 billion or many multiples more, Wu's more interesting bet is philosophical. In an era loud with predictions that AI will make engineers obsolete, he has staked his company on the opposite claim — that the best use of an autonomous engineer is to hand the toil to the machine and give the human back the joy of building things from nothing. It is a founder story still being written, by a founder who insists the humans should stay in charge of the story.
Your Story Could Be Next
Every founder in this series started exactly where you are now — with an idea and the will to build. What has changed is the leverage. Frederick gives you a team of AI agents that take you from idea to a real company, helping you design, build, and ship software on demand — no giant team and no endless runway required.
If Scott Wu's journey resonated, there is no better time to begin your own. Start building your company with Frederick.
References
- https://en.wikipedia.org/wiki/Scott_Wu
- https://techcrunch.com/2026/05/27/ai-coding-startup-cognition-raises-1b-at-25b-pre-money-valuation/
- https://techcrunch.com/2026/05/29/cognitions-scott-wu-says-ai-coding-agents-shouldnt-replace-humans/
- https://research.contrary.com/company/cognition
- https://www.lennysnewsletter.com/p/inside-devin-scott-wu
- https://colossus.com/article/scott-wu-tapes-cognition/
- https://cheekypint.substack.com/p/cognition-ceo-scott-wu-on-acquiring