Find the AI nobody approved.Learn more
Hacker News1d ago283 points48 commentslive
After 600 experiments, here's what I discovered about JEPA world models
After 600 experiments, a JEPA world model's success on long-horizon tasks improved from 22% to 80% by using hierarchical planning with 25-step subgoals, a method that is three times more effective than swapping in perfect physics.
Read the full story athaddyalchaer.comWhy this is in the Signal
LAXIMA AI Signal curates the highest-velocity stories across Hacker News, GitHub trending, and new Hugging Face / Replicate model releases — quality-filtered, deduplicated, and refreshed every four hours. This item surfaced from Hacker News with 283 points (by halcdev). We link straight to the original source above — see the full live feed.
More AI Signal briefs
- HNLovable apps expose passport details and more than 200K private messages
- HNTech Companies Are Trying to Shift the Concept of Privacy (Again)
- HNFM-1 Workbench
- HNMonkey Island: all 13 versions side by side
- HNTerminal borders that don't cost a cell
- HNLive airband on all frequencies from KORS. Victoria, CA also ground stopped
- HNJonathan Blow – If AI solved coding, why is my software still the same? [video]
- HNNvidia allegedly engineered scarcity by funding AI startups to buy its GPUs