https://xitter.cf/t.co/tN2bg09zl5
Germany
Joined July 2025
- Tweets625
- Following352
- Followers79
- Likes5.9K
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
just acquire Bend
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…
I think I'm affected by this. I like it so far. I will probably not like it when I hit my weekly limit in a few days though.
The difference between using Deepseek and using Opus is that you have to babysit Deepseek whilst Opus can just work.
introducing osplat, 1.8x faster training than gsplat (on my gpu)
if anyone wants this i can put it on github
wow, nice, who cares, next
GPT-6.1 Terra just surfaced in Microsoft's public code!
An EF Core update merged October 16 sets "gpt-6.1-terra" as the default judge across 18 evaluation configs. Seventeen previously used GPT-5.6 Terra. The same change moves the model performing the coding tasks to GPT-6.1 Sol.
The interesting detail: Terra is configured to grade another AI's work. Microsoft appears to be preparing its coding evaluations around a new generation of both models, although the configs don't establish a release date or reveal how much better Terra is.
