Will human code review still exist a year from now?
Hi HN, edgy thought, seeing a lot traction and usage of code review bots lately across both startups and medium/larger companies. Whether it's CodeRabbit, Graphite Diamond, etc. (there are plenty more).
If models keep improving, and we eventually inject these code review bots with more and more context to validate even things like business logic, tribal knowledge, etc. do you think we could eventually arrive very soon in a place where we no longer do traditional code reviews?
Just work with your coding agent to develop things, get it reviewed by a code review bot, fix any issues caught, and ship.
no, i just started doing copilot PR reviews on github and its pretty damn good.
unlikely- as long as your code is fairly micro-serviced about, llm generated code is fairly superior in all ways.
lol. human code review is all we will do day-to-day in a year
(i'm the lead engineer on diamond)
I'll still be reviewing my own code
I have tried getting LLMs to review code that I have written and most of the feedback I get is useless. It's as if they can only spot the most trivial of issues, or even worse -- they find issues in places where they don't exist.
I guess that they are moderately useful for finding copy-paste errors.
If i don't review the code anymore, dont expect me to be responsible for it.
Personally, I don't even think you'll have a code review bot. You'll just do it in the IDE.
A year? Yes, 10? Probably not.
I'm 100% confident LLMs could replace any code review being done by people who think this is a good idea, since both parties understand the assignment to the same degree.
Review bots are the autopilot. Humans are the pilots.
The autopilot will keep getting better, but you still want a human in the cockpit when something truly unexpected happens.