Borrowing an idea from one research discipline to solve a thorny problem in another is an age-old scientific trick. Charles Darwin’s theory of natural selection, for example, famously inspired algorithms in computer science that “evolve” solutions to complex problems. But such clever leaps are fleetingly rare. For #AI, however, hopping between fields can be a breeze. Researchers highlighted that ability with the release of 36 preprints produced with a single suite of AI tools called BootLoops. The so-called harness can be used with any large language model, such as Anthropic’s Claude or OpenAI’s ChatGPT. And in just 3 months, researchers used Claude outfitted with BootLoops to cross-pollinate at a massive scale, spotting—and precisely solving—problems in seemingly unrelated fields that share the same mathematical structures. The resulting papers range widely, from figuring out the age of sunspots, to analyzing government health policies, to attempting to decipher a mysterious medieval manuscript. Learn more: scim.ag/3TUUyDJ

We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals. Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale? Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run. The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range. We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward. We’ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!

GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration? DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it. That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders. With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets. openai.com/index/scaling-ai-…

SITUATION DETECTED: Demis Hassabis told the WSJ that his AGI timelines have shortened from three to five years away to within the next two to three years.

Gemini 4 Argon combined three open-source datasets to build an interactive 3D visualization of human anatomy. Toggle different layers in motion, isolate muscles, or zoom inside a beating heart.

Today we are excited to release Artios, a fully open source system for designing functional de novo enzymes. Enzymes present a new frontier for de novo protein design models due to the challenge of coordinating precise chemistry, rather than binding a target. Enzymes are life’s biological machines and by designing them we can truly start to engineer biology. We believe these fundamental capabilities should be built in the open and shared with the community to progress science faster. 🧵 1/6

Public preview of BioEmu2! @MSFTResearch AI for Science’s new foundation model for biomolecular structure, dynamics and function. Open weights, MIT license. github.com/microsoft/bioemu2 Movie: BioEmu2-simulated binding of intrinsically disordered peptide PUMA to MCL-1 receptor. 🧵⬇️

If we want to continue progress in science, we must be prepared to acknowledge that the correct description of nature will eventually simply be too complicated for humans to understand. I do not think it is wise to make human understanding a requirement. AI is the way forward.

LLMs can mimic thought and feeling with uncanny skill, but intelligence alone is not enough for subjective experience go.nature.com/3VSXWQ4

Fields Medalist who famously cried over OpenAI’s paper dump, refuses to use AI in his own research: “People that don’t want to use AI… they need to be able to practice their arts in the way they are doing it today.” the interviewer pushes back elsewhere: name one other field where we hold back technology that can already do the work, just for the sake of students. Hugo Duminil-Copin: “The human understanding of mathematics is the only thing that at the end benefits society. […] If the purpose of mathematics was purely getting as many results proved and lean formalized then […] you could even imagine a totally automatized version. […] You have a black box which is providing you the answer […] it gives you 42 but you have zero clue why.”