Vibe coding for teams, thoughts to date
9 points by pims
9 points by pims
I just had an epiphany about vibe coding. The definition of insanity is trying the same thing over and over again, expecting different results. The definition of programming insanity is trying the same thing (prompt) over and over again expecting the same results.
Well no. The LLMs output stochastically and are not determinstic. So, you can run the same prompt over and over again and get different results each time. For harder problems where the model has lower odds of successfully resolving the task you give it, you likely need to run it more times.
Git worktrees wind up being perfect for this. It’s basically rolling dice.
“bushy” is a good term for AI contribution. It’s usually not code that I as a human would want to read when added in large scale.
I think it would help a lot if it were possible to have git blame that says “written by AI” or something. That way I can at least know why what I’m reading looks so weird or doesn’t make sense.
agree, coding agents should have their own git identity, credientials, IAM roles and whatnot. “Sounds like work, not vibes” yup