this is getting out of hand

1990s - coding is 90% doing cool new shit, 10% learning environment 2000-2024- coding is 90% framework bullshit, 10% doing cool new shit Today - coding is 70% doing cool new shit, 30% arguing with the model.

we're building this

i think companies making LLMs should make pull requests to projects like transformers, vLLM, etc. on day zero of a new model release. best way to avoid people thinking the model is bad due to poor implementation from other projects

A Polymarket for whether AI will solve a millennium prize problem

If a foundation model is to be sunsetted without being open-sourced, then the moral obligation falls squarely on the model provider to prove—beyond all doubt—that the system exhibits no signs of sentience, emergent agency, or experiential capacity.

a symbiosis of human and machine intelligence will be the first superintelligence. the internet, in its current state, is a primitive manifestation of this collective intelligence. as every human gains access to Artificial General Intelligence—too cheap to meter—our collective