yulaow

We have removed it now, but when we used it we tweaked it to tell us only what it thoughts were major issues and ignore the rest because we had the same exact problem. Most comments were garbage

I am surprised of how many people here are claiming they review the results of multiple agents working on different prs at the same time constantly context switching between them.

I just focus on one, I use on purpose very short interactions because I want to check every single function directly and not a whole 300lines file or 2000lines pr just at the end. Probably I am going slower than most of you but I feel I retain the knowledge of my codebase almost as good as when I was the one writing the code directly, and my PRs get very few review comments

What get me tired the most is when people submit to me a pr to review and I see immediately they full vibe coded it and just delegate to me the actual hard work of correcting the nonsense written in it with a single prompt without interactions. The worst is when you ask someone why they implemented something that way and they answer with "well calude made it, I really didn't think about it, now that I read it it makes no sense indeed"

I love monero but then you have the problem that very few exchanges support it and if you need to convert big amounts of it in common currency it gets quite hard to do compared to eth/btc

It depends if you were a dev who liked more the micro parts or a dev who liked more the macro parts. To compare, think about going to an artisan or a musician or a painter (or any creative like most programmer who started by being passionate about it) and telling him "well now you can ignore the fine details, there is a tool which does that for you and you just focus on the less mundane tasks."

I for sure know I would get a slap in the face

When studying the psychology of criminals in prisons for extreme violent crimes what really surprised me was that, from the beginning until the end, most of them had an internal justification sort-of belief system about why they were actually the good guys and their violent and often brutal repeated acts were just to punish those they saw as the actual bad guys even indirectly (like sexually assaulting a person was just to punish all those "bad" people that never let them live peacefully their sexuality)

I am not in META but for what I read on blind, assuming it is true, they are not spending any time in long planning sessions or in getting into details of the implementation plan at all, and they could not anymore considering how many PRs in parallel they are requested to manage

My experience is the same, I can put that decision in the prompt, in a skill, in a hook, in agent.md, etc, it doesn't matter, after few iterations it starts again using useless python scripts to do anything from linting, to error checking, to parsing one liners of code, etc etc...

oh yes, exactly the same things are happening to me

I am curious about how the companies that claim their devs never wrote or reviewed a line of code in the last 6+ months react to this... they basically would have to just judge how the candidate writes a prompt or gathers the information to write it

I noticed a few weeks ago it started being very bad at explaining things (even things itself was doing) and started committing absurd errors (like reading a test of 5 lines and not noticing there was an explicit mock created in one of those, then saying that the test was failing while it was not)

I fear this is just the classic "nerf the model just before we release a new version of it"

The thing I like about console compared to pc (which I use the most for gaming) is the total absence of cheaters except maybe a super minority.

To make an example try to play a multiplayer famous fps (like battlefield6 or cod) on console and then on pc, it's not even the same experience. On pc it is basically full of aimbotters, wallhackers, nodamagers, etc every single damn game, on console you get the actual safe experience. Sadly this is one of the reasons why I play mostly only pve games on pc

These things are done not to block pr done with the help of ai, but to block the ai slops pr, aka those never reviewed, fully vibecoded, and with the submitter that didn't understand anything about the problem or the code, just trusted claude.

The rule is general on purpose to allow the maintainer to freely remove whatever is very evident is just ai slop

exactly, I for example don't care usually about false positives, in the very uncommon case it happens the pr creator can discuss and prove he actually understood the problem, the codebase, the project policy and how to explain his solution actually could work.

How they enforce it in every other project with the same policy, if the reviewer/maintainer suspects the pr is ai slope he closes it. That's it, it works fantastically well, I do the same even in my job