Should we just skip code review now?(xata.io)
xata.io
Should we just skip code review now?
https://xata.io/blog/ai-codes-humans-engineer
8 コメント
I would actually love to see this happen. Mostly for the potential schadenfreude.
Maybe.
Humans just can't thoroughly review the amounts of code AI can produce. But the answer is not YOLO and just ship blindly whatever AI created.
The answer is to move the human review/judgment upstream. You review specs. You review the steps. You review the verification rules. You review the contract the code must fulfill.
https://www.latent.space/p/reviews-dead
Humans just can't thoroughly review the amounts of code AI can produce. But the answer is not YOLO and just ship blindly whatever AI created.
The answer is to move the human review/judgment upstream. You review specs. You review the steps. You review the verification rules. You review the contract the code must fulfill.
https://www.latent.space/p/reviews-dead
My team are just vibe coding now and code review is non existent. Just hit approve. Why should anyone waste their time and brain cells reviewing a massive pile of slop. If management want us to go all in on AI then we will. Full trust in its output. I'm genuinely looking forward to the first major corporate disaster to happen because no one really pays attention to what code changes are being committed. When everything starts collapsing sell popcorn?
Aren't you at all invested in your team's outcomes?
Why would I be? If I bust my ass making sure all the vibe coded slop gets corrected the leadership will still give all the credit to AI and probably fire most of us because they think AI can do our jobs. Better to let it all burn if you ask me. If I'm going down then so are leadership.
no mention of scaffolding, upversioning old code, porting to new languages, refactoring. Alot of the things we talk about wanting to do are now greased.
a Still requires the engineering, but there seems , escaping the token mill leaves a powerful.
a Still requires the engineering, but there seems , escaping the token mill leaves a powerful.
Most AI slop was subsidized by $100 subscriptions. more limits will be added therefore the quantity of AI generated code will either drop or improve, just because writing code is "not free" anymore
but i do hope we will get to a moment where reviewing the code is not the bottleneck anymore (it has always been - just for different reasons)
but i do hope we will get to a moment where reviewing the code is not the bottleneck anymore (it has always been - just for different reasons)
We are largely now doing proof of functionality and a skim read and making sure it runs locally and deployed to a lower environment
All in all its still time consuming and we have discussed moving to a dark factory but there is still immense value in at least knowing what each file is for and maintaining good isolation of code