Cheeky Pint — August 27, 2025

Guest: Scott Wu (CEO, Cognition)
Host: John Collison ("Stripe")
Episode Theme: Scott Wu on acquiring Windsurf, AI replacing engineers, and the Moneyball-ification of everything

Episode Overview

In this candid, high-energy conversation, Stripe's John Collison sits down with Cognition CEO Scott Wu for a pint (Guinness for John; still no beer for Scott) to discuss Cognition's ambitious vision for AI in software engineering, their lightning-fast acquisition of Windsurf, the changing nature of engineering work, and how AI is re-shaping everything from job roles to economic infrastructure — all through the lens of Wu's distinct math-competition-honed worldview.

Key Discussion Points & Insights

1. Scott Wu’s Math Prodigy Roots and Founding Story

• Math beginnings: Scott grew up in Baton Rouge, inspired by his older brother and his parents’ engineering background.
  • Started doing math competitions at age 7, “competing in the seventh grade math division as a second grader” (02:12).
  • Early obsession with not placing: “I was just, I was so obsessed... That’s your supervillain origin story.” (03:06)
• Education & Early Career:
  • Left high school a year early; worked at Addepar before a brief college stint at Harvard (“I wasn’t that good at finishing school. I have a middle school degree... I didn’t really make it through high school or college.” [03:41]).
  • Close-knit group ties: Started at Addepar with now-prominent AI founders like Alexander Wang and Eugene Chen (04:11). “We all knew each other from middle school math competitions.” ([05:19])

2. The Age of Young Founders, Industry Maturity, and the “Moneyball-ification” of Everything

• Fewer “raw young” founders now: Scott argues being a founder’s gotten harder as the industry matures and there’s more “playbook” for company-building (06:44).
• Moneyball-ification:
  • As fields mature, intuition-based success yields to mathematical optimization (poker, chess, gaming, startups) (09:08):
    • “Now it's just all math nerds.”
    • Cites Super Smash Bros. and chess: “It’s like a difference in what skills get most selected for.”
• Notable quote: “There’s a lot of beauty in the nerd side of it too... It’s like a difference in what skills get most selected for.” — Scott Wu (11:28)

3. Cognition & Devin: The AI Junior Engineer

• What is Cognition/Devin?: "We’re building the AI software engineer... Devin, the agent; Windsurf, the IDE." (11:46)
  • Devin's paradigm shift: More than code completion (à la Copilot), Devin can be delegated actual tickets/projects, operating at a “junior engineer” level (13:34).
  • Real-world contributions:
    • Deployed across thousands of companies, including major banks.
    • “Devin is merging something in the range of like 30 to 40% of all pull requests that come through.” (14:32)
• Async vs. Sync Workflows:
  • Devin enables async delegation; future is merging async (agent) and sync (IDE/human-in-the-loop) workflows (15:26, 17:49).
  • Complexity distinctions: “The essential complexity is what software needs to do. The accidental complexity is all the routine stuff — and most of our time goes to the latter.” (15:26)
  • Biggest productivity jumps: Migrating, maintaining, refactoring — “when folks measure internally, they see something like an 8 to 15x gain for a lot of these use cases with Devin.” (19:27)

4. Productivity, Benchmarking, and AI Impact Measurement

• Hard to measure IDE productivity; agents make the ROI clearer:
  • “If anything, honestly, I think IDE productivity is often underrated...” (20:47)
• On AI “eating” coding tools jobs:
  • Scott: “I would kind of describe [the ‘labs just eat apps’ idea] as the nihilist computer use take... but in practice, there's a lot of contextual knowledge...” (22:04)
  • “Software engineering is so messy... that persists as an area where you need to specialize.” (24:09)

5. The Future of AI in Knowledge Work & AGI Perspectives

• Specialization persists:
  • Success is about “defining the right benchmark” for each task, expertise, and use case (24:09-25:52).
• On AGI:
  • “I think we have AGI.”
  • “You have to go define AGI... But my honest opinion is...we're just going to keep rolling out more improvements and these things are going to be more and more capable, but I don't know that we have some sudden shift.” (45:27-46:35)

6. The Lightning Acquisition of Windsurf

• Play-by-play recap:
  • News breaks on a Friday Google will acquire Windsurf IP/talent.
  • Cognition team “reached out cold that evening... handshake agreement by Saturday.”
  • Worked all weekend, “signed on Monday at 9am,” with public announcement soon after (46:43-49:04).
  • Strategic fit: Research/product stayed at Google; Cognition got enterprise/team/go-to-market. “We found a very natural fit.” (49:30)

7. AI Commerce, Economic Infrastructure, Agent Economy

• All-in on usage-based billing:
  • “Seats don’t really make sense when it is like the AI themselves are arguably seats as well.” (31:23)
  • “The other big one... is just for there to be an entire agent economy as well.”
  • Example: “We order our Doordash on Devin, we order our Amazon packages with Devin.” (32:17)
• “Bots allowed if you sign for them” world:
  • “I think what we will probably need to see a lot more of over time is basically like delegating access... an agent can do something on your behalf. And you are attaching some of your reputation to it too.” (36:11, 36:47)

8. How AI Shifts Hiring & the Nature of Engineering Work

• Smaller, founder-heavy team:
  • “Core engineering was about 19... now about 30 to 35.” — “Our whole interview process...is basically just having people build their own Devin in 8 hours and seeing how far they get.” (37:22, 38:15)
  • High-level thinkers, former founders: “21 of our 35 were founders before.”
• When will you hire your last engineer?
  • Scott predicts: “...two, three, four years from now where we stop using code as the main interface... Just instructing your computer and telling it what to do.” (39:08, 39:39)
  • “If anything, we will have way more software engineers, not fewer... We really never seem to run out of demand for more code.” (40:14, 41:07)

9. AI UIs and Product Lag

• Why have UI paradigms not changed yet?
  • Scott sees a decade of product innovation ahead, even if model capability gains stopped now (44:47)
  • “We could freeze all the capabilities today... and there would still be a whole decade of product progress to make.” (44:47)

10. AI Market Structure, M&A, and Industry Dynamics

• All layers win:
  • “I think all the layers are going to do very well... There’s just going to be a lot of AI.” (27:26)
  • Value accrues wherever there’s differentiation rather than vertical integration (29:06)
• M&A trends:
  • “...a little more polarized towards like you become a hyperscaler or bust.” (53:28)

11. Leadership & Learning Fast

• On learning to run a hypergrowth company:

  • “I’ve got a lot to learn still, for sure... you learn a lot from your peers and your friends who are doing similar things.” (55:40)
  • “It helps a lot to have a close group of friends. Then you can just be very honest...” (56:12)

• Information diet:

  • “Twitter is really for tech news, I think is really the place to be.”
  • “There’s a lot of video, but then I just don’t watch the videos.”
  • “I should have Devin... GitHub action the morning report...” (56:53-57:43)

Notable Quotes & Memorable Moments

• On what AI means for programming:

  • “We’ll get to a point where those theoretical concepts and that high level understanding... that is what programming will be. And if anything there’s going to be a lot more software engineers.” — Scott Wu (40:14)

• On the identity shift in engineering:

  • “Knowing all the little, memorizing all the facts or... the syntax... are going to be less important. What’s going to be more important is high level decision making [and] having a good intuitive sense of what to build.” — Scott Wu (38:19)

• On product opportunity:

  • “Honestly, if we factor in a lot of the more recent capabilities... it’s arguably less than one year for a lot of these. And we are all kind of grappling with that... trying to figure out what the right new product experiences are.” (44:47)

• On the “Moneyball-ification” of the world:

  • “At some point when the space gets mature enough... the conclusion of it kind of is math.” — Scott Wu (10:21)

• On AGI:

  • “Order your Doordash with Devin — sounds like AGI to me.” — John Collison (45:50)

Timestamps for Key Segments

• Scott’s math background & Addepar/AI founder connections: [01:17]–[05:43]
• Young founders vs. mature industries & Moneyball-ification: [06:36]–[11:41]
• What is Cognition/Devin, async/sync paradigms: [11:41]–[15:26]
• Devin in the enterprise, productivity impact: [14:29]–[21:37]
• Benchmarks, specialization vs. general AI: [24:09]–[26:55]
• Lightning M&A — Windsurf story: [46:35]–[51:41]
• AI’s impact on economic infrastructure, agent economy: [31:23]–[36:47]
• How AI changes engineering, Cognition’s hiring & interviews: [37:13]–[41:07]
• The coming decade of AI UI/product evolution: [42:07]–[45:26]
• Market structure, hyperscaler dynamics, and M&A: [27:02]–[53:28]
• Cognition's intense culture and M&A integration: [54:10]–[55:21]
• Leadership learning and information diet: [55:40]–[57:43]

Conclusion

This episode delivers a rapid-fire, insight-rich tour of both the transformative power and the messy realities of modern AI, told from the perspective of one of the most math-driven and hyper-ambitious new-generation CEOs. Scott Wu offers a simultaneously pragmatic and bullish vision of the agentic future — where software engineering is reinvented, agents buy your office whiteboards, and being a great engineer is about decision-making, not syntax.
Required listening for anyone building, using, or competing in the new AI economy.