I'm using claude but then refuse to do much cleaning up of what it spews. Im leaving that for the PR reviewers who love AI and going through slop. If they want slop, I'll give them the slop they want.

▲

whstl 9 hours ago | parent | next [-]

Not advocating that people should follow this but:

As someone that loves cleaning up code, I'm actually asking the vibe coders in the team (designer, PM and SEO guy) to just give me small PRs and then I clean up instead of reviewing. I know they will just put the text back in code anyway, so it's less work for me to refactor it.

With a caveat: if they give me >1000 lines or too many features in the same PR, I ask them to reduce the scope, sometimes to start from scratch.

And I also started doing this with another engineer: no review cycle, we just clean up each other's code and merge.

I'm honestly surprised at how much I prefer this to the traditional structure of code reviews.

Additionally, I don't have to follow Jira tickets with lengthy SEO specs or "please change this according to Figma". They just the changes themselves and we go on with our lives.

	▲	grvdrm 6 hours ago \| parent [-]
		Favorited. I was talking to someone (non-dev) yesterday who prototypes with Claude and then goes back/forth with the lead engineer to clean it up and make it production worthy (or at least more robust). I like that model.

▲

MSFT_Edging 8 hours ago | parent | prev | next [-]

Just started work on a project. Greenfield and "AI accelerated". PRs diffs are in the range of 10s of thousands of lines. In the PR, it is suggested to not actually read all the code as it would take too long.

▲

jmccaf 8 hours ago | parent [-]

If you push a change, or you approve, you're responsible for the change and its effects later. Regardless of size. If change is too big, tell your teammates its too big to review and to refactor to bite-size with their great coding agents. Use AI models also for review of large changes, consider a checklist . Setup CI and integration tests (also can be AI assisted)

	▲	JTbane 3 hours ago \| parent \| next [-]
		Agreed, and something will go wrong (as every junior has experienced). You cannot lay blame on the AI when git blame shows your name.
	▲	MSFT_Edging 5 hours ago \| parent \| prev [-]
		Oh there's plenty of CI, linting, etc. Half of which is not properly plumbed in.

▲

kelzier 9 hours ago | parent | prev | next [-]

I thought the de facto policy was that the individual remains responsible in a team context.

	▲	notpachet 9 hours ago \| parent [-]
		Sweet summer child.

▲

jbxntuehineoh 8 hours ago | parent | prev [-]

based. our CEO has made it clear that we're expected to use LLMs to shit out as many features as we can as quickly as we can, so that's exactly what I'm doing. Can't wait to watch leadership flail around in a year or two when the long term consequences start to become apparent

	▲	red75prime 7 hours ago \| parent \| next [-]
		> when the long term consequences start to become apparent Choose your own story! and then a) programmers become relevant again and slowly fix all this crap, b) Claude 7.16 waltz through fixing things as it goes.
	▲	eloisius 8 hours ago \| parent \| prev [-]
		You'll just get laid off and they'll be onto the next hype cycle as visionaries.