AI news, research and bad takes. accelerationist. professional shitposter in training per aspera ad astra
Joined March 2024
- Tweets494
- Following754
- Followers71
- Likes2.1K
Mino retweeted
Just been locked out of my account and forced to put paid partnership tags on two of my posts that weren't paid partnerships on Kimi K3.1
These were tested in arena before it even released and I have had no affiliation with Kimi / Moonshot, and have received no early access or payment from them, I have seen a lot of other people getting this so what tf is going on ?
I even came out and called out Minimax for doing similar the same thing offering paid partnerships without tags (I declined those) this is very unfair as I didn't even test the model after it released only in arena on the 22nd of October a day before release !
@elonmusk @allegrajacchia @Nate_Esparza
they’re literally ramming security escorts off the road to steal millions in server racks
we’re doing GTA missions for datacenters now
❗️ Cargo thieves have stolen more than $100M in hardware bound for US data centers since June, NBC News reports.
They hijack freight-broker accounts, post fake loads and send carriers malware links to take over their accounts too.
In one of the biggest heists, an SUV struck the security escort of a trailer carrying ~$10M in server rack components from California to Texas.
The trailer sped off. An app was tracking it, but its coordinates had apparently been spoofed.
In another, a sedan hit the off-duty police escorts of a $13M server load bound for Georgia, and the trailer drove on.
None of the equipment from either load has been found.
Mino retweeted
People seem to forget how unprecedented this is
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
Mino retweeted
Assuming their claim holds, E^(NP) does not have linear size circuits!
See Thm 6.1 in cs.cmu.edu/~ryanw/improved-a…
some context on how important these results could be
the Exponential Time Hypothesis is one of the main assumptions about the fundamental limits of computation. If it’s false, hundreds of results about how hard certain problems are would need to be reconsidered, potentially opening the door to much faster algorithms.
the math result would settle three major conjectures connecting wave behavior, geometry and harmonic analysis, some of which have remained unsolved for decades. All dimensions, with a single new proof technique.
the physics result could finally provide rigorous mathematical foundations for some of the quantum field theories we use to describe fundamental particles and their interactions, something physicists have been trying to establish for decades.
none of this means immediate technological breakthroughs, and the proofs still need independent examination.
but if they hold up, we’re looking at an AI system potentially resolving decades-old problems across three completely different scientific fields in a matter of weeks
OpenAI just announced that its AI models have produced proofs for three major open problems across mathematics, computer science and physics
- proved the local smoothing conjecture in all dimensions, which would also resolve the Kakeya and Stein restriction conjectures
- disproved the Exponential Time Hypothesis, an assumption underlying hundreds of computer science papers
- produced rigorous existence proofs for 4D Yang-Mills theory and QCD
the math and ETH proofs have been formalized in Lean, along with the Yang-Mills existence proof. some of the world’s leading researchers, including two of the original authors of ETH and a Nobel Prize-winning physicist, are now involved in examining and explaining these results.
the wildest part is that OpenAI is describing the AI-generated proofs as “raw research data” that human scientists need to turn into scientific understanding
OpenAI just announced that its AI models have produced proofs for three major open problems across mathematics, computer science and physics
- proved the local smoothing conjecture in all dimensions, which would also resolve the Kakeya and Stein restriction conjectures
- disproved the Exponential Time Hypothesis, an assumption underlying hundreds of computer science papers
- produced rigorous existence proofs for 4D Yang-Mills theory and QCD
the math and ETH proofs have been formalized in Lean, along with the Yang-Mills existence proof. some of the world’s leading researchers, including two of the original authors of ETH and a Nobel Prize-winning physicist, are now involved in examining and explaining these results.
the wildest part is that OpenAI is describing the AI-generated proofs as “raw research data” that human scientists need to turn into scientific understanding
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
openai.com/index/building-sc…
OpenAI is clearly positioning GPT-6.1 Terra around visual reasoning and computer use
understanding what’s on screen, taking actions across apps, and handling everyday work
big upgrade over 5.6 Terra, just not a new frontier model
Mino retweeted
GPT-6.1 Terra by @OpenAI is now in the Arena. Head to Agent Arena to test it out, scores coming soon!
In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks from a global community of users. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.
GPT-6.1 Terra is also available in Code Arena: WebDev, Text, Vision, Search and Document.
OpenAI says GPT-6.1 Terra now matches GPT-6 Astra on Agents' Last Exam at 1/10th the price per token
and finishes tasks over 3x faster than both Astra and GPT-6.1 Sol
this is their mid-tier model
Anthropic's AMD compute agreement includes a commitment by AMD to buy up to $5B of Anthropic stock if certain conditions are met
chip suppliers are increasingly becoming investors in the companies buying their compute
Anthropic estimates that 5–7% of its revenue in the first nine months of 2026 came from customers it identified as violating its usage policies or commercial terms
Anthropic says Claude wrote more than 90% of the contributions to its own codebase in September 2026
and it's now helping build the systems used to train and evaluate the next generations of Claude
"𝘵𝘩𝘦 𝘢𝘷𝘦𝘳𝘢𝘨𝘦 𝘥𝘢𝘪𝘭𝘺 𝘷𝘰𝘭𝘶𝘮𝘦 𝘰𝘧 𝘤𝘰𝘥𝘦 𝘤𝘰𝘯𝘵𝘳𝘪𝘣𝘶𝘵𝘦𝘥 𝘵𝘰 𝘰𝘶𝘳 𝘤𝘰𝘥𝘦𝘣𝘢𝘴𝘦 𝘱𝘦𝘳 𝘦𝘯𝘨𝘪𝘯𝘦𝘦𝘳 𝘪𝘯 𝘵𝘩𝘦 𝘵𝘩𝘳𝘦𝘦 𝘮𝘰𝘯𝘵𝘩𝘴 𝘦𝘯𝘥𝘦𝘥 𝘚𝘦𝘱𝘵𝘦𝘮𝘣𝘦𝘳 30, 2026 𝘸𝘢𝘴 𝘰𝘷𝘦𝘳 45 𝘵𝘪𝘮𝘦𝘴 𝘵𝘩𝘦 𝘢𝘷𝘦𝘳𝘢𝘨𝘦 𝘥𝘢𝘪𝘭𝘺 𝘷𝘰𝘭𝘶𝘮𝘦 𝘪𝘯 𝘵𝘩𝘦 𝘺𝘦𝘢𝘳 𝘦𝘯𝘥𝘦𝘥 𝘋𝘦𝘤𝘦𝘮𝘣𝘦𝘳 31, 2024"
insane
one of the stranger things in Anthropic's S-1 is that "model welfare" could be a reason to delay retiring older Claude models
AI moral status is now an actual risk factor in an IPO filing
Anthropic delayed the start of pre-training for a new frontier model by more than a month to improve security against model weight theft
one month of delay at the frontier just to secure the weights
