Science & medicine

Essays on AI in science: genomics, medicine, mathematics and formal proof, materials, chemistry and research integrity.

It Output the Numbers They Said It Did

18 minute read

· Science & medicine

A Nature Sustainability paper estimates that 403 US hyperscale data centres used about 82 TWh in a year and emitted 25.7 million tonnes of CO2. In the previous review round the same electricity produced 44.6 million tonnes, and the paper said these facilities ran on dirtier power than the US grid. A reviewer who read the code against the methods section found the denominator was wrong. After the fix, the comparison points the other way.

Not in the Register, Already in the Literature

16 minute read

· Science & medicine

A Google researcher set 48 AI agents loose on public NASA lidar and announced a Maya site that Mexico’s heritage institute had no record of. Archaeologists had already published it. Spotting the site was never the hard part, and a register that doesn’t list a place was always weak evidence that nobody knew about it. I went through the blog post, the papers it cites and twenty years of lidar studies to work out how much β€œnot registered” actually tells you.

Ask for 300 Ideas, Get Sparse Autoencoders 128 Times

28 minute read

· Science & medicine

A NeurIPS 2026 paper splits a research idea into two scores: whether it hangs together, and whether anyone in the field is placed to think of it. Ranking for the first and against the second gives a broader set of ideas than asking Claude or Gemini, with the same quality per idea. I recounted the released files, went through the code and the hosted production version, and set out how to use it, which setting to choose, and what the evidence can and can’t support.

Three Streams in One Genome

77 minute read

· Science & medicine

The ancestry app gave me one letter for my father’s line and crashed on my mother’s. So I read my raw 30x genome myself: a maternal line in India for about 50,000 years, a paternal line (J2a1-L26, down to the J-Z6090 branch) that points to the Iranian plateau, and a measured mix of about 33% First Indian, 52% Indus-related and 15% Bronze Age Steppe ancestry.

Twenty-Seven of Thirty-Three Left Blank

14 minute read

· Science & medicine

Two Microsoft researchers ran a 100-minute research sprint with twelve researchers in three teams and AI tools on every desk. Observers logged 33 AI interactions. One was accepted outright, five were modified, none were rejected, and 27 never got a recorded decision. The blank column is the finding.

Zero Rotatable Bonds, Published in 2014

12 minute read

· Science & medicine

DeepMind’s cheminformatics demo has an agent retrace the crizotinib-to-lorlatinib rigidification in about an hour, work the post says normally takes days. The pair it lands on has been in the literature since 2014, and the headline relationship is two PubChem fields. What the demo leaves untested is the part I would pay for.

The Old Network Names More Lesions. The Question Is Which Ones It Sends Home.

10 minute read

· Science & medicine

A dermatologist tested a 2015 image network and two frontier language models on 200 biopsied skin lesions. The network named the right diagnosis far more often. On the decision a patient cares about, whether a lesion can be sent home, the comparison only means something once the two are set to the same level of caution, and at its default setting the network discharged 18 of 30 melanomas.

Twice, Four Times or 1.7 Times: The Multiple Depends on Which Doctors

9 minute read

· Science & medicine

Microsoft’s diagnostic orchestrator is now in Science, and the coverage says it is twice as good as doctors. The same post also says nearly four times and 1.7 times. All three are true, and each is a different comparison: the physicians had references or did not, and the cases were NEJM puzzles or ordinary emergency admissions.

Solved on the Set It Was Scored On

13 minute read

· Science & medicine

A DeepMind researcher says AlphaFold 2 scored about 90 on its test set and that this does not mean protein dynamics are understood. Four 2026 evaluations show where the gap sits: models lean on whichever state the Protein Data Bank has more of, and one benchmark of a membrane transporter puts an AlphaFold 3 model 0.57 angstroms from its outward-facing structure and 5.42 from its inward-facing one.

OpenAI’s First Math Errata: Three Papers Outside the Lean Catalogue, and One Corollary the Lean Never Covered

17 minute read

· Science & medicine

A day after releasing 722 AI-written mathematics manuscripts, OpenAI withdrew three of them over a sign error and revised fourteen more. None of the three withdrawn papers is in the release’s catalogue of Lean formalizations. Of the thirty manuscripts the update touched, only one is in the release’s catalogue of formalized papers, and its correction landed in a corollary the formal proof never covered.

OpenAI’s 722 Math Papers: The Proofs Can Be Machine-Checked. The Statements Are Still on Us.

16 minute read

· Science & medicine

OpenAI released 722 AI-written mathematics manuscripts on 6 October, including a claimed proof that the Riemann zeta function has no zeros to the right of 7/8. The numbers live in the GitHub repository, not the announcement. 162 papers come with a formalized main result. That is a real check, but only of the proof. Whether each Lean statement says what its paper claims is a separate question, and the repository’s own catalogue lists its review status as β€œunchecked”.

Which Word in the Headline Has Been Measured?

11 minute read

· Science & medicine

Vals AI’s post reports two room-temperature magnetic semiconductor candidates found by Claude agents. The accompanying ledger is unusually frank: one compound has never been made and may not survive being made, the other was made in 1999, and no gap or spin measurement exists for either. The words of the headline have different evidentiary status, and only candidates is fully supported.

Claude Found the Enzyme System in One Campaign Out of Eleven

18 minute read

· Science & medicine

Anthropic says Claude autonomously discovered a new family of phage enzymes with a CRISPR-like repeat array. Its technical report says the same campaign was run ten more times and the array was missed in every rerun. Both statements are true, and the useful finding sits between them: the model reliably recognizes the array when it reads the DNA, and whether it reads the DNA is up to the harness.

The Advisory Group Got Four Written Guarantees. The One Thing It Can’t Advise On Is the Pace.

23 minute read

· Science & medicine

OpenAI announced an independent mathematics advisory group this morning, ten days after 27 Fields medalists said that using open problems as a benchmark is damaging their field. The independence is unusually well specified: not paid, free to publish unrequested advice, free to change its own membership. Then one closing sentence removes internal pace from the remit. Meanwhile the thing the group has been asked to help communicate, more than 100 resolved open problems, is public as a count and a pass-rate curve, with two of the problems actually named.

page 1 of 2 · 23 essays on Science & medicine