Weekly AI for Science newsletter extracts cross-field patterns from top research
bravo_abad · x · 2026-09-21
Substack author Jorge Bravo Abad runs Discovery at Scale, a weekly newsletter distilling the ideas behind each week's AI for Science papers, focusing on patterns that repeat across fields and lessons that transfer between them.
The latest issue covers tumor immunology work showing that training on experimental function improves prediction of tumor-reactive T-cell receptors. It uses the case to discuss a recurring scientific pitfall: when the quantity you truly care about is hard and expensive to measure, it's tempting to substitute an easier proxy — and the piece analyzes how to handle that proxy-vs-target gap properly.
Related event: AI for Science: New TCR Prediction Method and Weekly Methodology Digest(2 posts)→
More from AGI Musings
- The Zvi on why AI risk isn't solvable: "Someone else builds Skynet" — TheZvi · 2026-09-21
- Air-gapped PCs leak data via heat, magnetic fields and FM radio — 1 bit/hour to 40 bits/sec — StefanoGogioso · 2026-09-21
- A dad let ChatGPT choose his car — the B2A channel for big purchases is real — jeff_weinstein · 2026-09-21
- Nathan Lambert: Chinese open-weight models lead US by 1.6B downloads and months on benchmarks — Interconnects (Nathan Lambert) · 2026-09-21
- Grady Booch: Contemporary AI is better called statistical intelligence — blaizedsouza · 2026-09-21
- Compiler analogy debate: is vibe coding automation or losing the loop? — binarybits · 2026-09-21