Post

Short musings on "cognitive debt" - I'm seeing this in my own work, where excessive unreviewed AI-generated code leads me to lose a firm mental model of what I've built, which then makes it harder to confidently make future decisions simonwillison.net/2026/Feb/15/...

So is the conclusion that we can't vibe code large projects because we can't maintain them? Has anyone managed to keep vibe coding at scale?

My conclusion is that you can only go so far not understanding how the project you're building works - you may not need to look at every line of code, but you gotta slow down and make sure your mental model is accurate enough to stay productive

This 100%, but (in these wild times there's always a but) you can ask the robot "how does this codebase work?", followed by "why doesn't it work like the following: ...", and it will happily refactor the entire thing. So I feel that even in 100% vibe mode, you can retake control *if* you want.

I’ve been working with it a lot and I’ve got two habits that seem to help. One it atomic commits per feature with a good commit message and the other is getting the agent to add/update a mental model doc which is kept intensionally brief. This then allows me to both load up my brain and the AI’s