I think we’re getting a @OpenAI operating system next week at Dev Day. Linux based, and if I were to guess probably built on Arch with a Gnome fork, or possibly even a fully custom DE. Arch to be at the leading edge and take advantage of all the money and effort already being thrown at it by Valve, Gnome-like to appeal to MacOS familiarity with their 1 billion users.
Fully malleable agentic operating systems are obviously the future, and the natural home for powerful coding models. @OmarchyLinux was the first to fully embrace it and garner meaningful public attention, but it’ll be far from the last. If we get Steam Machine and OpenAI OS in the same year, then 2026 is definitively the year of the Linux desktop.
I took the time to read it, and there’s little I disagree with. The academic community seems to be focused on very different things however, understandably centred around how they can continue being academics, and what the value of their research will be without the massive compute of frontier labs.
Those of us on the other side of the argument are excited for a world where the hardest problems in mathematics don’t go unsolved for 80+ years. I want progress, and whether that progress comes from humans with pen and paper, or from a tool that humans built is less important than the work itself. It’s also completely possible that some problems might be outside of human ability on any timescale due to sheer breadth of scope. Like waiting for a chimp to build a combustion engine. Screw that. Let’s build.
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: https://t.co/RuEosScSMb
We’re at an interesting inflection point where you’re almost equally likely to be called a dumbass for not believing that super human intelligence is about to kill us all, or be called a dumbass for thinking that AI shows any form of intelligence whatsoever. The balance is almost beautiful.
We’re working with an independent advisory group of mathematicians to help OpenAI responsibly share advances in AI and mathematics.
The group will advise on how we assess and communicate new mathematical results, uphold academic and professional standards, and build tools that support mathematical research and learning.
Through this work, we want mathematicians to be at the center of shaping how AI supports mathematical understanding and how its benefits reach the wider community.
https://t.co/QCMFLFlIv8
We’re working with an independent advisory group of mathematicians to help OpenAI responsibly share advances in AI and mathematics.
The group will advise on how we assess and communicate new mathematical results, uphold academic and professional standards, and build tools that support mathematical research and learning.
Through this work, we want mathematicians to be at the center of shaping how AI supports mathematical understanding and how its benefits reach the wider community.
https://t.co/QCMFLFlIv8
@scheemunai You’re talking about intelligence/$$$ ratio, and leaving out the intelligence part of the equation. How many GPT 4o tokens would it take to solve Navier-Stokes? Infinity?
I assume you, like me, also learned math from examples, and didn’t derive it from first principles. It sounds like you’re arguing that current AI models are not *very* intelligent because the amount of examples they need to see to learn something is vastly larger than what a human needs. That’s a fair criticism of the current state of learning algorithms, but that’s different from a claim that these models categorically have zero intelligence.
“The primary goal of mathematics has always been about human understanding of the ideas.”
Respectfully, I disagree. Perhaps that’s the primary goal of mathematicians, but not of mathematics. Math doesn’t have goals, it just is. Don’t claim there are grave dangers at hand because the pace of progress is going to outstrip the mathematical community’s ability to digest it. Math isn’t under threat here, but rather the mathematicians can see the door closing on being the last great explorers of the mathematical universe, and are understandably sad at not being able to put their name on some small island in it. AI is about to hand them a map and a compass to something so vast that it exceeds the ability of anyone to comprehend all but a small region—but that’s where we were already, and the addition of having tools for navigation sounds to me like the opposite of “grave risks and dangers”.
The real danger to me is having some of the brightest humans on the planet spending their lives toiling away at some small corner of mathematics only to either never have a breakthrough, or for it to be of so little consequence for humanity that only a handful will ever take the time to understand or make use of it. Let’s instead encourage AI venture ahead and create the map for us, and we can choose what to explore.
https://t.co/nHX0YW58DG via @YouTube