Cognition
Its agent fixed the founders’ server on Christmas Eve. The brand still splits its aspiration in two.
The buried story
On Christmas Eve 2023, a small team in a New York apartment was stuck on a broken server. They handed the problem to the AI agent they'd been building.
It found a faulty system file the team had missed, deleted it, configured the server and finished without asking for help. Scott Wu called it the first time Devin handled a real task on its own.
That moment came after eight pivots. It's the best founding story in AI coding, and it sits in interviews instead of on the website.
Where Cognition stands
Wu, Steven Hao and Walden Yan hold ten gold medals from the International Olympiad in Informatics between them. Wu placed first in the world in 2014. In the company's words, they "built the product we always wanted for ourselves."
Devin launched in March 2024. Annual revenue went from $1 million to $73 million between September 2024 and June 2025, with total net burn under $20 million. Cognition bought Windsurf for about $2.4 billion, and Founders Fund then led a $400 million round at a $10.2 billion valuation.
Goldman Sachs, Citi, Dell, Cisco and Palantir run Devin in production.
The brand line: "Devin works like a capable, collaborative teammate."
Why the line falls short
"Teammate" describes a product. Wu has already written two bigger ideas: engineers becoming architects who direct agents, and the software that never gets built because developers cost $100,000 to $300,000 a year.
Both ideas appear in the funding post, in back-to-back paragraphs. Neither leads the brand.
Friction (8.0). Engineering capacity is the bottleneck, and code completion only moves it. Linktree's month with Devin produced about 300 pull requests. Copilot and Cursor name the same problem, though.
Lived Stake (8.5). The highest in the series for AI coding. Elite programmers built the tool they wanted, and Christmas Eve is the proof.
Distinctive Qualification (8.5). A 73x revenue ramp on little cash, blue-chip customers and a team no rival can copy. Devin still has to prove it can replace senior judgment.
Real Aspiration (6.0). The weakest score, and the fix is already written. It's split across two claims that haven't been joined.
The story Cognition should tell
On Christmas Eve, stuck on a server we couldn't fix, we gave our agent a chance and it solved the problem. Engineers will design the systems and agents will build them, so the software that was always worth making finally gets made.
Three moves to raise the score
1. Join the two aspiration claims. One statement that pairs engineers as architects with the software that never got built. Wu should write the long version. Real Aspiration: 6.0 to 8.5.
2. Put Christmas Eve on the front door. A short founder story on the About page: the apartment, the pivots, the broken server. Lived Stake: 8.5 to 9.0.
3. Show the economics. Most of Cognition's revenue is already margin positive, which is rare in AI apps. Make that as visible as the revenue growth. Distinctive Qualification: 8.5 to 9.0.
If all three land, Cognition moves from 7.8 to about 8.6.
