Jul 18, 2025 · 51m · 20vc
Cognition CEO Scott Wu on Acquiring Windsurf: The Process, The Deal, The Rationale · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the 20VC podcast, host Harry Stebbings interviews Scott Wu, co-founder and CEO of Cognition, about the strategic acquisition of Windsurf, the evolving capabilities of AI-driven developer workflows, and the competitive dynamics of the broader artificial intelligence industry.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 22.6% of the talking time here. How this is scored →
Harry as informed peer 5.5 Guest teaching 4.4 Guest disagreement 2.5 Harry pushing back 5.1
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 27:17 Scott pushes back on Lovable growth comparisons
Scott directly counters Harry's claim that competitors are outgrowing Devin by revealing Cognition matched Lovable's revenue trajectory over the same period.
Hardest push from Harry ▶ 24:36 Harry challenges Devin's narrative lossHarry refuses to accept polite PR answers and directly tells the CEO that his product has fallen out of the industry zeitgeist.
Biggest teaching moment ▶ 27:17 Scott corrects Harry on Cognition's actual ARR growthScott exposes Harry's faulty premise regarding Devin's market traction by sharing non-public growth metrics, forcing Harry to state 'I stand corrected'.
Harry holds his own ▶ 13:02 Harry challenges application layer defense vs AnthropicHarry uses venture capital logic to question why foundation model providers wouldn't vertically integrate and capture the application layer's margins.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Origins of the Windsurf Acquisition | 3 | 2 | 1 | 2 | Harry asks standard exploratory questions about the origin of the Windsurf acquisition. Scott details the fast Friday-to-Monday transaction timeline in a collaborative tone. | |
| Rationale and Execution of the Transaction | 6 | 4 | 2 | 6 | Harry pushes back on the deal rationale, asking why Windsurf was not just a deteriorated 'husk' like Scale, and citing VC opinions on whether Google made a blunder. Scott gently reframes the asset value around existing customers, IP, and team fit. | |
| The AI Talent War and Elite Researchers | 5 | 4 | 3 | 5 | Harry interrupts Scott's discussion on talent compensation to demand a specific numerical quantum of elite AI researchers. Scott offers a contrarian defense of high talent compensation while narrowing the elite researcher count to between 100 and 1000. | |
| Model Providers and Application Layer Value | 7 | 4 | 4 | 8 | Harry presses hard on Cognition's dependency on Anthropic, challenging why foundation labs wouldn't simply steamroll the application layer and asking for revenue split percentages. Scott holds his ground, arguing for layer differentiation and declining to disclose revenue share figures. | |
| Reinforcement Learning and coding Jevons Paradox | 6 | 5 | 1 | 4 | Harry demonstrates domain expertise by citing recent interviews with Marc Benioff and Vlad Tenev regarding 50% AI code generation. Scott explains reinforcement learning progress and Jevons paradox in software engineering. | |
| The Future of AI Agents and Deep Context | 6 | 5 | 1 | 5 | Harry performs live ROI math comparing $300k developer salaries to Devin's hourly pricing to test value capture. Scott elaborates on deep context and how developer roles will shift toward technical architecture. | |
| Enterprise Multipliers vs Consumer Attention | 7 | 6 | 5 | 8 | Harry directly confronts Scott with the premise that Devin fell out of the zeitgeist compared to Cursor and Lovable. Scott counters forcefully by revealing unannounced revenue figures matching Lovable's hypergrowth, prompting Harry to concede his premise. | |
| The Future of Next-Generation Human-Computer Interfaces | 5 | 5 | 3 | 5 | Harry attempts to force Scott into naming Cognition's primary competitor or common enemy. Scott sidesteps by quoting Jensen Huang on finding non-zero-sum paths to victory. | |
| Blending Synchronous IDEs and Asynchronous Agents | 3 | 4 | 1 | 1 | A brief collaborative discussion where Scott outlines the product roadmap for combining local IDEs with asynchronous coding agents following the Windsurf acquisition. | |
| Quick Fire Round and Concluding Thoughts | 7 | 5 | 4 | 7 | Harry conducts a fast-paced quickfire round, pushing Scott on deal structures, foundation model consolidation, and critiquing Cognition's launch video. Scott defends his market projections and accepts PR advice while maintaining emotional composure. |
Statements from this episode (19)
Wu: The decline of founders going down with the ship is disappointing
“I think there's an unspoken covenant that as a founder, you go down with the ship, and I think that for better or for worse, it's changed a bit over the last year, and I think it's a bit disappointing, to be honest.”
Wu: Current AI coding capabilities alone would fundamentally transform the world
“In AI code, to be truly honest, if there was zero progress, the world would still be entirely different.”
Cognition learned of Windsurf's Google deal at the same time as the public
“We found out, honestly, on Friday, the same time everyone else did, that all of this was happening, and that this was the split, and that it was Google”
Cognition initiated cold outreach to acquire Windsurf on a Friday night
“We reached out cold Friday evening and our first conversation was Friday night.”
Cognition negotiated and signed the Windsurf acquisition over a single weekend
“Get, get to a verbal agreement on Saturday, hash out all the details and the terms and the legal on Sunday, get everything signed and done on Monday morning.”
Wu: Fewer than 10,000 researchers drive the vast majority of AI progress
“Somewhere in between those two is right. You know, I think there are at least a hundred folks that, that, that, that are, that really matter in making a lot. A lot of differences. Certainly less than 10,000 and probably a good bit less than that, if I had to g…”
Wu: AI application layer talent competition is fiercer than at foundation labs
“I think the level of talent and the fierceness of competition is, in many of these is actually like, Greater even in the application layer, which is crazy to say, because I think the competition in the Foundation Labs is extremely strong.”
Wu: Reinforcement learning is AI's biggest breakthrough of the last 18 months
“You know, it's RL, for example, is I would say the biggest breakthrough of the last year and a half call it.”
Wu: AI tools currently give software engineers a 1.5x to 2x speedup
“And I would say that something in that range of, you know, 1.5 to two X feels right to me today in aggregate.”
Wu: AI will make software engineers 10x faster within three years
“Look, I think in three years, there's no reason that shouldn't be a 10 X.”
Cognition prices Devin hourly to be 10x cheaper than developer time
“So it's all usage based, but it's essentially by the hour. And we try to make things so that they're about 10 X cheaper than you know, than basically the value of your time and going into it.”
Wu: AI will make 30 million global software engineers 10x more efficient
“There are thirty million software engineers in the world. You know, we're going to make them all 10 X more efficient over these coming years. We're going to be writing 10 X more code.”
Wu: Cognition's Devin usage grew 5x to 10x in the past six months
“In the last six months, you know, between January and now, you know, like, even aside from this, you know, obviously the latest deal this week with Windsor The usage of Devon has actually grown something like five to 10 X. And that's been the case in both self…”
Wu: AI agents will match IDE familiarity within six to 12 months
“I think you fast forward six to 12 months from now, and I think people will be familiar with agents to the same level that, that they're familiar with IDEs today.”
Cognition added roughly $80M in run-rate revenue in six to eight months
“Yeah, look, I don't want to get into exact numbers, but even with the windsurf deal aside, like not including that at all, we are basically, we've done roughly that as well in the last six to eight months.”
Wu: Future software engineering will replace coding with expressing intent
“At some point, telling your computer what to do is not going to take place with code. It is going to take place with you just expressing your intent.”
Wu: Private AI foundation labs will far exceed $500B aggregate valuation
“If you add all their valuations today, what is that? Like, 500 B, something like that? I think that's gonna go way up in the next five years.”