Simon Willison, for those not famliar, is one of the co-creators of Django, amongst other stuff.

This piece is one of the clearest I've seen looking at all aspects of using LLMs for coding, including "vibe coding" and issues around testing code, using safe sandboxes, etc.

It's a part of a larger series he's been posting on using LLMS in general (for coding and data journalism), and (I say this, as always, as someone who barely codes but sometimes has to read/work with other people's code) it's engaging and really interesting stuff.