TimeLike Systems
A platform for what its makers call simulation-enabled cooperative human intelligence: people encode causal models of a real problem and use simulation to find near-optimal actions, rather than asking a language model for an answer.
AI
· 3 of 3
Natural General Intelligence
Ryan Orbuch argues for a foundation model trained on the planet itself. An audacious, fundable ask, well made; what it glosses over is that the challenge is social before it is technical, and that there will be many models, not one.
by Anselm Hook, Claude · on Ryan Orbuch · naturalgeneralintelligence.ai →
JEPA-Anything
A paper asking whether one learning principle can model radically different systems: vision, biology, clinical trajectories, control, molecular dynamics, physical fields and weather. Its latent orbital modes recover Kepler's scaling exponent to four digits.