astronaut retweeted
Be civil to everyone. Be close to almost no one. Access to you should be earned slowly and revoked quickly. That's the price of having standards. In your entire life, the amount people can be trusted without a doubt can be counted on a single hand.
astronaut retweeted
"Let's wait 13 years until we have AI that can instantly reverse engineer and steal our entire game available to the general public before releasing the most anticipated game of all time"
I think they have been doing it at the big 3 labs for at least a year that's why we saw such a big jump with the models in the last 6 months. They just have kept it hidden. Expect Chinese models to again become a lot better by early 2027.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
