Aananth V (@aananth_) profile picture

Grad student @UTAustin. Previously @Google.

Austin
Joined January 2022
Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)Photo by Aananth V (@aananth_)
So when are they releasing Math 2?
11
I don’t understand the hype around Neon? The model is good, but it is certainly not as good as Opus 5.5.
28
Honestly, all these problems exist with GPT 6 Astra as well. Haven’t tried 6.2, but wouldn’t be surprised. The Sol models have been much more “usable” in my experience. But no GPT model has come close to Opus 5.5.
89
Can a formally verified C compiler be as fast as gcc? CompCert’s optimizations come with a machine-checked proof that they don’t change your program’s semantics. The catch? The compiled code runs ~28% slower than gcc -O2. We tried closing that gap with AI agents 🧵
8
23
3
121
11,160
Where we landed, vs gcc -O2: - CompCert benchmarks: 28% behind → 4–6% - PolyBench loop nests: 55% behind → ~14% - Held-out larger programs: 13% behind → ~9% On the first two, we’re now faster than gcc -O1.
3
1
20
833
Who tf decided that the stopwatch notification should show the lap button and not the reset button
77
I think it's time to switch to Claude, Codex has been a slop machine lately
1
72
Is Google back?
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model. Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards. As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!
1
130
Terra is draining usage much faster than yesterday
39
Antigravity slowly becoming my favorite IDE for doing research. Switching between Argon and 3.8 Flash feels very seamless and I can say "Argon, what do you think?" right in the middle of the conversation!
1
1
78
Please fix the memory leak @thsottiaux
2
50
Even the withdrawal README is written in astra-speak ugh.
Based on a hot tip: I had Astra check the OpenAI paper "Algebraicity of Weil classes on split abelian eightfolds", and it seems to be flawed. This is one of the nonformalized results. Concerning.
1
2
270
I just hope that super intelligence is benevolent
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math
35
Thinking about writing a formally verified programming language for Kernel generation, think we're not far from /all/ software being written in provably correct languages.
3
73
Can someone at @OpenAI un-RLHF the "It is X, not Y or Z" behavior from Sol and Astra?
1
88
Aananth V retweeted
the fly brain can play beat saber
1,051
8,014
2,431
85,248
23,138,659
Any rational actor will stop working on the doomsday machine
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
59