The papers and results moving the field forward: scaling insights, new architectures, reasoning and multimodal progress, and the benchmarks that measure them. Every digest links the paper, the code, or the lab's own write-up so you can go straight to the source.
Axiom Math says its AI agents formally verified the Stacks Project — about 16 million lines of Lean code — betting verification, not slowdown, keeps math credible.
Discovered Materials, a months-old YC startup, is offering $200,000 for each of five open materials science problems.
Potential Sciences, a UK startup launched Nov 5, wants to simulate molecular physics at scale instead of running wet-lab experiments, betting synthetic data can train drug-discovery AI where public datasets fall short.
Denise Herzing is pairing decades of dolphin audio and video with AI from Georgia Tech and Google DeepMind — and testing two-way sound exchanges underwater.
MIT Technology Review convenes editors Mary Beth Griggs and Antonio Regalado to examine how technologists are pursuing continuity of the self past death.
Today's Download covers "reverse game theory" as a lever for agent control, plus the Trump administration's "Super Intelligence" rebrand and a roundup of tech stories.
Alignment researchers are turning to mechanism design — 'reverse game theory' — to make rule-following the rational choice for AI agents, sidestepping the need to inspect model internals.
Two critical PHP bugs sent teams offline to patch. Wiz's CTO says AI-accelerated discovery means this is a preview, not a one-off.
Nathan Lambert argues that building meaningful frontier benchmarks is harder than improving benchmark scores.
