Post

I am currently in a state when I would like to use more agents type of work (works fully automated) for bugfixes at least. but I am "afraid" it will make some shit. I review most of the fixes even if its do a full test with real servers and real data.

16/10/26, 8:15 pm0 reposts3 likes

Anthony Panozzo

@panozzaj.com

Maybe one way I think about it: what would give you confidence that the changes were correct / good / maintainable / etc.? Then build in or work on those sorts of checks, bug report logic, code reviews

Chris Arter

@arter.dev

I do this for a large codebase at work. Make your quality gating done via scripts that are enforced on precommit hook and in the CI. dependency cruiser to enforce architecture, strong lint, & minimum test coverages. no "no mistakes" prompts. actual deterministic gates arter.dev/posts/the-an...

arter.devThe Antidote to Code SlopHow Determinism-in-the-loop saves your tokens, and your sanity.

Sergii Yarochevskyi 🇺🇦

@aft1n.bsky.social

You're saying you get more slop from agents then real production-grade code? Or its just more of the attitude?