AI-assisted coding for teams that can't get away with vibes

24 points by mogambo1


simonw

This is genuinely excellent - I think it’s the best piece I’ve read on AI-assisted development patterns.

I partially like the emphasis on how robust software engineering practices help AI tools perform much better - comprehensive and well designed tests, linters, CI, continuous deployment - things that make humans more productive make AI assistance more productive too.

pyj

A system in which AI thrives is one with markers of a high quality team and codebase.

This is likely because the messier codebase was as confusing for the AI as it would be for a human.

This seems like a continuation of the theme I see where AI often exposes existing inefficiency in software development. A lot of the article about directing AI to complete tasks works just as well if you s/AI/junior software developer/g.

migurski

Rhymes strongly with this other story: https://lobste.rs/s/ldfeqs/gap_through_which_we_praise_machine

The engineers’ skill is required to prompt well.

benji

They have a keen calibration and feel for what leads to a good system and can steer LLMs accordingly, i.e., they have what I like to call “the mechanic’s touch”.

A perhaps more common term for this is “mechanical sympathy”. (Often misunderstood by car enthusiasts as not abusing the machine, but that misunderstands the use of “sympathy” here which is more about deep understanding.)