Vibe-code with stable infrastructure

Vibe-coding produces close-to-unmaintainable code right now. But it is an acceptable trade-off in places where it produces decent results and maintainability matters less -- for example making simple UIs.

So maybe a good approach is to vibe-code with stable infrastructure. This riffs off of Facebook's engineering principle "move fast with stable infrastructure", itself updated from the original "move fast and break things".

That is, make the core APIs good and have reasonably LLM-readable documentation for them. And then rip through the UIs with Cursor. Make a lot of them, quickly, for every purpose that is needed. The main bottleneck will then be talking to users, understanding the problem, and making sure what gets built is a reasonable solution -- as opposed to coding.

This kind of approach seems like a great fit for building internal tools. These are typically operations and admin-related web apps, and are an established category of software in most organizations beyond 10 engineers or so. It is also unglamorous work, so doing this with maximum AI seems especially appropriate.

I find this application of vibe-coding exciting because most potential internal tools have never been built, because having engineers do it is not worth the opportunity cost (hence Retool's success). But if you can now vibe-code them 10x or even 100x faster, there could be 100x more.