"after 11 technical interviews" "your work has been accepted into our codebase" interviews are getting out of hand

unpatched infinite money glitch: if claude can't solve your problem you can repackage it as a reinforcement learning environment and sell it to anthropic

so many gen Zers are like "the world used to be so colorful" no dude you just grew up watching cocomelon-style, colorful cartoons, which over-saturated your memories of how things actually looked

the duo sold out in 5 minutes... insane

YC startup: it's too complicated, we'll use a mix of claude and chatgpt to rebuild it from first principles

we're really living in the timeline that built AGI on jupyter notebooks and trained it primarily on javascript

rockstar waited too long to release gta6 because claude kept better so they kept using it to build more features but now fable can just zero shot gta6 in a couple of hours probably

the problem with ai is that now we get even more open source slop weekend projects

can someone who's built a dark software factory with agents tell me a bit about it

which mac mini is that btw and can i have it for my cluster?

sometime ago i had posted that even though anthropic can build slack internally overnight they don't because it's not worth the effort, so they just use slack well they built a new internal version of git, built around agent swarms merging their code, because git was designed for human work, and it became a bottleneck for agents. when (not if) slack becomes a bottleneck for agents it's next on the chopping block. and i think that's how most software will get phased out. agents doing increasingly more autonomous work will be limited by the next bottleneck in the existing software pipeline, which will just get optimized away, by a tool built for agents, not humans. most software will just run headless and we'll just see the final output