The inevitability does not come from whether or not AI is at our mercy, but from the fact that there are human groups (which are not at our mercy) that can use it as and when they want.
@dotcsv
-
Reflection on superintelligence’s first manifestations in mathematics
By
–
Let us live this moment with the transcendence it deserves, for I believe we are witnessing one of the first obvious manifestations of what an artificial superintelligence can offer us. For now, it is a fact limited to the field of mathematics, we do not
-
Anthropic mathematician praises AI math breakthrough
By
–
Alpoge (the Anthropic mathematician who was linked to the Navier-Stokes controversy) applauds these results and emphasizes that this is one of the most important moments in the history of mathematics.
-
OpenAI releases 372 math problem results on GitHub
By
–
🔴 OPENAI RELEASES ADVANCES IN MATHEMATICS! Finally, OpenAI has published the hundreds of math problem results it had been keeping under lock and key in recent weeks. It is 372 results that they published in a GitHub repository, covering advances in many
-
AI model competes with popular Chinese models during ongoing post-processing
By
–
The model competes head-to-head with the most popular Chinese models, in a training that is still continuing its post-processing, with which it could still gain more capabilities.
-
Mistral announces Large 4, 1T multimodal model
By
–
🔴 ¡ANUNCIADO MISTRAL LARGE 4!
— Carlos Santana (@DotCSV) 6 octobre 2026
Vuelve el gato gordo de Mistral 😽 La compañía europea ha anunciado hoy su nuevo modelo de 1T, multimodal y con un rendimiento que lo sitúa a la cabeza de los modelos abiertos europeos (no era dificil) y de USA https://t.co/6YLvs9eGXQ🔴 MISTRAL LARGE 4 ANNOUNCED! Mistral's big cat is back 😽 The European company today announced its new 1T model, multimodal and with performance that places it at the top of European open models (that wasn't hard) and those of
-
AI and scientific credit: Claude’s discovery and human authorship
By
–
> On AI and scientific credit The authors of the article are, in this case, human, and it is they who assume all responsibility and credit for this work. However, the discovery itself comes from Claude and an internal Anthropic employee. This is where
-
Claude achieves scientific result in algorithmic complexity
By
–
Regarding AI and scientific discoveries, something curious happened today. First, the news: Claude has produced another fairly important scientific result in the field of algorithmic complexity. This paper demonstrates having surpassed a
-
Comparing LLM speed by tokens per task, not tokens per second
By
–
It makes no sense to compare models by speed (tok/sec) when they require different amounts of tokens to solve the same task. If you want a fair comparison, then measure their time per task, and there you will see that since GPTs generate far fewer tokens than Claude, their
-
ChatGPT/Codex gets 50% speed boost on Astra 6 and Sol 6.1
By
–
Oh wow, Tibo promised within the next 28 days an improvement to ChatGPT/Codex that would be good enough, or otherwise he'd give us a RESET. Well, today he kept his word: a 50% speed improvement on Astra 6 and Sol 6.1, just like that, just for fun. It's in