Sherwin Wu V2
If you talk to a lot of engineers, the thing that annoys them the most is after you've written your beautiful code, how do you get it into production, covering engineering tradeoffs, AI product work, and team leadership.
Episode
Sherwin Wu V2
Summary
Sherwin Wu, an engineering manager at OpenAI, discusses how AI coding tools like Codex have transformed software engineering — 95% of OpenAI engineers use Codex daily and 100% of PRs are reviewed by it. The conversation covers how engineers are becoming "tech leads managing fleets of agents," the counterintuitive prediction that AI may usher in a golden age for B2B SaaS, and the challenges of deploying AI tools broadly across non-engineering organizations.
Key Takeaways
Engineers who use AI coding tools heavily are opening 70% more PRs than those who don't — the productivity divide between early adopters and laggards compounds quickly.
Don't over-index on customer feedback when building AI products: models evolve so fast that they will "eat your scaffolding for breakfast," making yesterday's feedback obsolete.
To drive real AI adoption inside a company, you need a dedicated internal tiger team that explores the technology bottoms-up — top-down mandates without this infrastructure rarely stick.
The "one person billion dollar startup" narrative has a second-order effect: enabling that requires hundreds of other small startups building bespoke software, which may actually expand the B2B SaaS market.
When a 100% AI-written codebase breaks, you lose the escape hatch of fixing it manually — teams must invest in tests, monitoring, and clear ownership to operate without that fallback.
Notable Quotes
“I think the same thing might happen for people management as well, especially in tech companies. And we're already seeing this. There's some teams where there are EMs managing quite a few people and they're doing it pretty adeptly because of some of these tools where they can get higher leverage and understand what their team's doing, understand organizational context a little bit better and operate in that way.”
“One, you're kind of starting to see this play out in the AI startup scene where software's became a lot more vertical oriented, where these verticals, like creating some AI tool for some vertical tends to work quite well because you really lean into that particular domain, you really understand the use case for it. And so if you play out AI, there's no reason why you can't have like 100x more of these startups.”
“One of these principles is what I talked to you about before, which is spending a lot of time with top performers, like actually spending... And to be very concrete, it's like more than 50% of your time with your top performers, with maybe your top 10% performers, and really, really trying your best to empower them.”