someone make a kid from astra and fable
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…
Same Python code, different tokenizers: 253 tokens with GPT-2, 138 with cl100k_base. That's 45% fewer tokens without changing the code. The biggest difference? Whitespace. cl100k_base merges indentation runs, as a result gpt-4 was really good with python compared to gpt-2.
Learnt about tokenizers and built my own over the last 3 days and it’s so cool!
github.com/flcrom/tokenizer
checkout tiktokenizer.vercel.app its such a cool website if youre learning about llm and stuff
Built my own version: a ferrofluid surface that rises toward a nearby magnet using just the phone's magnetometer. Making this a wallpaper for my pixel next.
Vardhman retweeted
private keys from 8 intermediate CA certs have been compromised (eMudhra)
Happy "Update everything" day to all my pager-carrying friends
emudhra.we-have-a-problem.co…
Pretty good idea.
Funnily enough, I broadly agree with Mr. @DeGatchi. For vulnerabilities like this one which impacts everybody using KVM (read every hyperscaler, every AI lab, every company on the planet really), there should be a fund that pays from all of them.
As Vercel we are trying to do our part here. And all credit to @PaulosYibelo for going down the pass of responsible disclosure!
😭😭
you can now take your frustration out on roadmaps
Press SHIFT + R to try it out at roadmap.sh/frontend