AI4S NotesThe subject is the meeting point between a learned model and a scientific claim. The notes follow that connection through proteins, molecules, materials, physical simulation, Earth systems and experimental practice.
A model can compress an expensive calculation, suggest a candidate or organize a search. The interesting question is what evidence would justify the next step. This notebook returns to experimental controls, distribution shifts, uncertainty and the work needed to turn a promising prediction into useful knowledge.
Original notes, traceable references
The essays are original, AI-assisted editorial writing prepared for this site. They do not reproduce source articles or claim new experimental results. References lead to the research that provides context. The opening collection was first published on 7 November 2026.
A weekly research watch
The watch collects selected recent items from public arXiv, Nature Machine Intelligence and Google DeepMind feeds. Only work connected to scientific applications is eligible. The short reading prompts are written for this notebook. External titles are bibliographic references, and the papers remain with their original publishers.
Updates run weekly. A feed may show only its latest daily edition, so a selection is never presented as a complete account of the preceding week. If a source is unavailable, its last available data can remain visible with an update-status notice.
A quiet, static site
There are no accounts, comments, submission forms, advertising scripts or analytics cookies. Reading does not require registration. Links to outside publishers are clearly marked. No personal profile, contact address or social account is published here.
Every page is served as a static document. The weekly publication process runs separately and has no public administration interface.