llms won't solve cancer by themselves wrote about what's missing from the current AI for science discourse and why we need more than just intelligence was an absolute pleasure to work with @swyx on this piece!

don’t type pymol commands urself, now there’s co-pymol available for claude code and cursor ⬇️

Fable 5 is live on shortcut.ai anthropic.com/news/claude-fa…

barring the fact that i do really need a better calendar app, love what @fiftyyears is doing

the work of saints <3

can we call them venture post-capitalists

i made The Norman Polycule up >:) behind-the-scenes, highlights from the 2,000 applicants, etc, here: rawandferal.com/norman-polyc…

Agreed there is no single modality in biology that is as information-complete as text. Bio ML (and imo specifically the Virtual Cell space) has a very real metrology problem. But there are many problems that ML in biology can solve when you take key bets on specific modalities and applications. ML in bio is presently useful for reducing combinatorially large search spaces with closed loop signal from wet lab verification. Whether or not "foundation models" help is a function of whether the data modalities that were scaled encode useful priors for a given downstream task. > @newlimit is taking a bet on epigenetic reprogramming and train models that predict cellular aging from TF perturbations. Yamanaka factors are POC that certain combinations of ~1.6k human TFs can magically reprogram cells. > Polyphron (@FabioZB_I) is training models to search over combinations of perturbations (small molecules, mechanical, electrical) for tissue differentiation > @chaidiscovery's Chai-2 reports 16% hit rate on de novo antibody design vs <0.1% prior > Antibody Fc region design and mRNA design are both verifiable problems where the functional assays provide clear signal and hypothesis generation can be optimized The difference between reducing large search spaces and solving tasks zero-shot is a function of model capabilities, which improves with more data and wet-lab feedback. Scaling "works" in the sense that these architectures can learn any objective function we throw at them.

(1/5) Great to be at @sequoia to give a sneak peek of one of our research directions! TL;DR one path to data-efficiency may be to “abuse GPUs like they’ve never been abused before”

Take it from me, guys, you got to spend your arrogance while you're still young. You're gonna spend all your 20s trying to get there, and in your 30s you realize there is no there. You're going to realize that the skills that get you to the party aren't the skills that will get you to be happy with the rest of your life. You're going to go from quitting smoking to smoking again to quitting smoking in one day. You know what? You're just not smoking right now and that’s fine. The only constant will be time spend on the codex app will probs increase.

Who actually shapes AI policy in the U.S.? We mapped 1,812 entities: 745 people, 918 organizations, 2,925 relationships. Frontier Labs, AI Safety orgs, Think Tanks, Government, VCs, and more. mapping-ai.org

Applications are now open for the 2026 Paradigm Fellowship: a 4-day retreat for young people who are obsessively good at something technical. For our fourth year, we're expanding to welcome builders across every frontier — AI, robotics, energy, bio, prediction markets, or something we haven't thought of. Last year's cohort came from 10 countries. Some were undergrads, some were dropouts, some were founders, some came from OpenAI, SpaceX, Citadel, and Kalshi. The format is simple: firesides, whiteboarding sessions, and time to hack. What makes it special is what happens in between, and after. Fellows have met cofounders, started companies, and gone on to raise from Paradigm and others. I was a fellow before joining Paradigm, the retreat was a transformative trip for me, and I met some of my closest friends through the program. Apply by June 8th. Retreat runs August 12–15th.

Introducing MTS: The first timeline-native news network that's always on. Monitoring tech, finance, geopolitics and culture — as it happens. We are Live Now.

Ok, it's time

Mythos system card came out today. A step change in ability, saturating many of the leading SWE benchmarks, novel heights at cybersec, and much better at math also. I read through and wrote down some highlights:

still waiting for the rest of the files but @donaldjewkes did such an incredible job retelling this saga!

Memo: What If We're Right? We recently wrote a private letter to partners & friends of a common failure mode: the inability to consistently reason through the daisy chain of downstream consequence when non-consensus, low-probability, events actually occur pages: 1-3

🚨 We're open-sourcing Druids, a library for coordinating and deploying coding agents across machines. Our beta users have used Druids to work on open math problems, conduct ML "autoresearch," and make software faster.

Excited to release PostTrainBench v1.0! This benchmark evaluates the ability of frontier AI agents to post-train language models in a simplified setting. We believe this is a first step toward tracking progress in recursive self-improvement 🧵:

they're all leaving 🤔 monitor-the-situation.com/mi…