Yet another opinion on LLMs · Hasufell's blog

The approach I’ve ended up on is essentially “pair programming”. Discuss approaches with it, come up with a plan, let it code it if you want, then look it over. Got a bug? It’ll suggest tests and investigations; if you think it’s missing something you can tell it. Just adds a huge amount of value this way. On the other hand, if you just tell it “go implement this feature” like a manager, it might not do so well. Possibly a multi-agent approach where you have one instance doing the work and another reviewing and the two iterating might work: I’ve heard of people doing this but I haven’t tried it.