When code generation is nearly free, the bottleneck moves to trust: does this diff actually do what it claims, and who's accountable when it doesn't. We hit this building an AI-native product last year: our CI bill dropped but our review/rollback tooling spend tripled, almost the inverse of the old ratio. Do you think Replit's all-in-one model wins because it collapses that trust gap, or does it just move the risk somewhere less visible?
The way I think about it is: GitHub is about managing who does what, Vercel is about getting it live, and Replit is about making sure it actually works. Each one is betting on a different bottleneck now that writing code isn't what takes up most of our time anymore.
It looks like AI generated content. I don't know why include stripe to this article
Embedded ad 🤷♀️
As usual insightful write-up.
When code generation is nearly free, the bottleneck moves to trust: does this diff actually do what it claims, and who's accountable when it doesn't. We hit this building an AI-native product last year: our CI bill dropped but our review/rollback tooling spend tripled, almost the inverse of the old ratio. Do you think Replit's all-in-one model wins because it collapses that trust gap, or does it just move the risk somewhere less visible?
The way I think about it is: GitHub is about managing who does what, Vercel is about getting it live, and Replit is about making sure it actually works. Each one is betting on a different bottleneck now that writing code isn't what takes up most of our time anymore.
Do you think all three directions are equally good bets or are there is a winner here?