On June 4, Simon Willison picked up a thought from Charity Majors that I haven’t been able to shake for days. It’s about the friction between AI enthusiasts and AI skeptics — two camps that often sit on the same team and both want to build great software. Majors’ image for it: the enthusiasts are in a race against time, the skeptics in a race against entropy.
Both Sides Are Right — and That’s the Point
The enthusiasts are not wrong. We’re seeing real, discontinuous leaps in capability from teams that lean hard into working with AI. It doesn’t feel like a normal technology cycle where you can calmly wait for the dust to settle. Teams that sit it out while competitors hustle could be out of business before anything settles at all. A real, existential threat.
But the skeptics are not wrong either. When you ship code faster than humans can read it, in domains nobody fully holds in their head anymore, you’re making withdrawals from a trust account that took years to build. Reliability degrades, institutional knowledge evaporates. What’s left are systems nobody understands, products burbling into incoherence, and on-call rotations that grind people up. Also a real, existential threat.
That’s the uncomfortable thing about this debate: there’s no easy side to pick. Both risks are real, and they pull in opposite directions.
The Missing Link
Majors’ actual punchline, which Willison highlights: there is no natural feedback loop connecting enthusiasts with skeptics. The enthusiast feels the speed and the opportunity. The skeptic feels the creeping debt. But the two signals rarely meet — they talk past each other, often in the same standup.
Majors doesn’t frame this as a fight one side has to win, but as a leadership and engineering challenge: how do you deliberately design feedback loops that ‘mend the gap in shared reality’ between the two groups?
My Take
That framing is what makes the difference. The usual AI argument runs as a camp war — you’re Team Hype or Team Brakes. Majors flips it: both hold a true part of the story, and the failure is the missing bridge between them. I’m the type who runs near the front myself. But the most dangerous sentence in any team isn’t ‘that won’t work’ — it’s ‘nobody really understands it anymore, but it runs.’ Force enthusiasts and skeptics to actually listen to each other, and you end up with more robust software than either side builds alone. That’s not a tooling problem. It’s organizational design.
Sources: Simon Willison: AI enthusiasts are in a race against time, AI skeptics are in a race against entropy, Charity Majors: AI enthusiasts are in a race against time