Pinned Tweet
GPT-6 Astra is the new #1 on BoardBench.
It scores 67.2 vs the previous SOTA Gemini 4 Argon at 66.3.
But it costs 2x as much ($0.56 per puzzle vs $0.28) for less than one point!
Check it out at boardbench.dev
1
5
1,031
okay i'm learning magic tricks now so that when @elonmusk solves money i have something fun to do
2
5
180
the one and only @AdamHoltererer
OpenAI doubled rate limits. Is it enough? I broke down the value of what you get right now, how it compares to before, and how it compares to the Claude Code subscriptions
6
283
eliancode retweeted
AA's time per task completely ignores:
- # of tool calls
- ttft latency between tool calls
- how long the tool calls take to run
2
1
20
2,196
Insane stuff.
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…
Community note
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
1
56
wait you're telling me this wasn't a thing before?💀
Message threads are now in public beta on the Messages API.
The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
3
336
