Academic and industry teams reported major advances in AI for small‑molecule discovery and biochemical simulation. MIT and Recursion released Boltz‑2, an open‑source model that predicts ligand binding affinities in seconds and topped CASP16 benchmarks; SandboxAQ and other groups described large quantitative models (LQMs) grounded in physics for molecular simulation. GEN covered these developments and their potential to compress early discovery timelines. Boltz‑2 demonstrated orders‑of‑magnitude speed improvements over free‑energy perturbation and came with permissive licensing to enable industry adoption. Parallel announcements from major tech labs described multi‑billion‑parameter foundation models trained to propose and prioritize candidate molecules and combinations. The rapid demonstration of predictive binding and physics‑aware LQMs suggests AI is moving from hypothesis generation to quantitatively useful lead prioritization, lowering costs for screening and enabling smaller biotechs to compete on discovery speed.