I've been thinking more about the this reply and I think I figured it out. > how did you trust it before when humans wrote it? you mostly didn't, not really. you trusted the tests, the fact that it was already in production somewhere similar, the author's track record, and the reality that most of the time the bug would show up in monitoring before it became a real problem. the diff was honestly a ritual more than a guarantee. we pretended like line-level review was a safety net because we didn't have better options and the volume was low enough that it felt manageable. but agents just make the old ritual obvious. once you're landing thousands of changes a month the idea that someone is carefully reading every diff starts being theater. trusting agent output is the same problem as trusting human output, just with higher throughput. I think the answer is still to stop treating the change as the thing you trust and start treating the environment as the thing you trust. if an agent can run the code, drive the ui, check the invariants, and get blocked by the same constraints a human would, the fact that it wrote the patch matters less. every time you find yourself wanting to open the diff "just to make sure," that's signal that something in the system is still relying on human eyeballs instead of a hard constraint. the code review as we know it is already on the way out. most of what we call review today will turn into risk scoring and continuous verification. low and medium risk work will land after agents and checks have already exercised it. humans will still be involved, but not by reading patches. they'll just be deciding what outcomes matter, whats not allowed to break, and which constraints are worth making permanent. the job now is to continuously make the system good enough that you don't need the diff to feel safe.

whoever lets agents in gets the volume. whoever blocks them will pay for it later. someone should solve this btw. not me though.

if any ai frontier company actually believed they were building superintelligence this is what they would do.

refusal tax has to be paid. dax is right you need agents going 24/7 but how can you do that if they tap out the minute you have to touch "security". a model refusing mid incident is a massive liability. fuck claude. i seriously hope every refusal is a churn event.

the flaw with git is becoming obvious and it slows autonomous agents down. git was built for humans taking turns. put a bot team on one repo and the bottleneck is collisions. someone's actually gonna build source control for agents and it's gonna be huge. not me though.

I cnt believe that worked. Our industry is under yet another paradigm shift. Goodnight

Im running an experiment where I let grok bots fully build a game from scratch in godot. I haven't seen the code. idc. No reading the diff. i just play the game. If it's fun we keep going. If its not fun we go back to the drawing board. It's two lanes of 1 engineer bot, 1 reviewer bots, 2 tester bots working in parallel. So far we have 876 commits, 240 PRs merged, ~12 merges/hour, 42 merged in one hour at peak, 22k lines of gdscript. This will either be awesome or be the most incredible slop we've ever seen.

me telling my grok bots "No fucking blocking!"

i read through quite a bit of this last night and rape allegations aside, i have some thoughts. if i recall correctly i've spoken about how i have a problem with most fictional writers' prose. it's one of the reasons why i tend to lean non-fiction. i just have a hard time finding a fiction book that reads well. well after reading this i figured it out. it's not only the prose but it's that most fiction writers suck at telling a story. i'm reading fiction to be entertained. when i'm reading non fiction the purpose is usually knowledge. and if i'm expecting to be entertained then i expect you to be able to tell a good story. but most fiction writers do not care about telling a good story. they're just up their own ass and like playing with words. i tried reading dog stars and every now and then i come back to it but it's a slog because the author doesn't want to write a story, he wants to write poetry. with the terror, sometimes i wonder if dan simmons is going for a personal best. are you writing this because you're trying to prove to yourself that you can? what's the point of all of this nesting? why are you pulling away from the story? i'm not interested in your game or your opinions. i don't care about you trying to be clever and hiding details to see if your readers would catch it. i don't care about your shared universe or callback. i don't care about you settling a score with the culture. fuck off and tell me a good story. i only have dnf fiction books because these fiction writers don't care about storytelling. they care about their ego. fiction writers love wrapping their stories in complexity and flowery language because it serves their ego. they want you to know they're a master of words. unfortunately you don't prove that to me by your prose being overwrought and your story being convoluted. this is a good story. the prose is good because all of it serves to tell a good story. it reads how someone would actually think and talk. if wbe was in a room with me he could physically tell me the story and i'd be just as entertained. if there happens to be a deeper message in there it's not because he chose to write it in, it's a side effect of his good storytelling. he's just writing what he's experienced. it's honest and whether it's real or all in his head it doesn't matter. it's still his experience.

are skills like the new dotfiles?