AI4S Noteshttps://notes.ai4s.co.uk/Original AI for Science reading notes and a weekly research watch.enSun, 08 Nov 2026 01:00:18 +0000Research watch · 08 Nov 2026https://notes.ai4s.co.uk/watch/2026-11-08/https://notes.ai4s.co.uk/watch/2026-11-08/This watch brings together 24 recent source items, with a particular presence in molecular science, life sciences, earth & climate. The links are starting points for reading, not a ranking of scientific importance.Sun, 08 Nov 2026 01:00:18 +0000What turns a prediction into a discovery claim?https://notes.ai4s.co.uk/notes/prediction-and-the-discovery-claim/https://notes.ai4s.co.uk/notes/prediction-and-the-discovery-claim/A useful prediction changes the next experiment. The strength of the discovery claim depends on what that experiment can rule out.Sat, 07 Nov 2026 04:27:44 +0000A dataset split is a scientific experimenthttps://notes.ai4s.co.uk/notes/dataset-splits-as-experiments/https://notes.ai4s.co.uk/notes/dataset-splits-as-experiments/The test set defines the kind of unfamiliarity a molecular model must handle. Choose that unfamiliarity before comparing scores.Sat, 07 Nov 2026 04:27:44 +0000Reading a protein structure as a functional hypothesishttps://notes.ai4s.co.uk/notes/protein-structure-and-function/https://notes.ai4s.co.uk/notes/protein-structure-and-function/A predicted structure can organize an experiment without settling what a protein does in its biological setting.Sat, 07 Nov 2026 04:27:44 +0000The distance between a promising material and a usable samplehttps://notes.ai4s.co.uk/notes/materials-prediction-and-synthesis/https://notes.ai4s.co.uk/notes/materials-prediction-and-synthesis/Candidate ranking is only one decision in a synthesis workflow. A useful evaluation follows the material into the laboratory.Sat, 07 Nov 2026 04:27:44 +0000What a physical constraint buys a learned surrogatehttps://notes.ai4s.co.uk/notes/surrogates-and-physical-constraints/https://notes.ai4s.co.uk/notes/surrogates-and-physical-constraints/A small equation residual is evidence about one test. Reliable use also depends on boundaries, regimes, and accumulated error.Sat, 07 Nov 2026 04:27:44 +0000Why weather skill needs a separate climate argumenthttps://notes.ai4s.co.uk/notes/weather-skill-and-climate-claims/https://notes.ai4s.co.uk/notes/weather-skill-and-climate-claims/Short-range predictive skill and long-term climate claims require different evidence, even when they describe the same atmosphere.Sat, 07 Nov 2026 04:27:44 +0000Keeping controls inside an AI experiment loophttps://notes.ai4s.co.uk/notes/closed-loop-experiments-and-controls/https://notes.ai4s.co.uk/notes/closed-loop-experiments-and-controls/An adaptive system needs a record of its decisions and controls that survive adaptation. Otherwise, iteration can outrun inference.Sat, 07 Nov 2026 04:27:44 +0000Research watch · 07 Nov 2026https://notes.ai4s.co.uk/watch/2026-11-07/https://notes.ai4s.co.uk/watch/2026-11-07/This watch brings together 24 recent source items, with a particular presence in molecular science, life sciences, earth & climate. The links are starting points for reading, not a ranking of scientific importance.Sat, 07 Nov 2026 04:27:44 +0000