Loading activity…
@haozhangml
Asst. Prof. @UCSanDiego running @haoailab. ex: @meta Superintelligence Lab, @snowflake (aquihire), @lmsysorg, PhD @scsatcmu
AI
A user estimates self-hosted fastH3-omni at about $0.14 per 15-second 768p clip, versus about $11.31 for Seedance's 1080p output.
AI
A user shared the @haoailab link with three rocket emojis and no explanation.
AI
Arena says its index tracks unauthorized actions, false attribution and deceptive completion across 90,000-plus agent sessions and 27 models.
YouTubeAI
A user says JEV-like decision models are pretty good at playing real-time games, but that they found a lot of problems during development. They recommend trying the demo.
AI
A user poses the combination as an open question, without naming a specific application.
AI
A post says a distillation run greatly improved quality and teases a bigger system-side release for the week of September 21, 2026.
AI
Posts say 25 Fields Medal winners signed a declaration warning of a mismatch between AI companies and mathematics. Supporters stress understanding and scientific communities; critics say AI proofs can still yield insights.
AI
The team reports the result on eight Nvidia B300 GPUs, generating 1344×768 video with stereo audio. Timing excludes model loading, compilation warmup and final MP4 encoding.
AI
Announcement enables local generation of h3-level clips on MacBooks and DGX systems.
AI
Official account highlights public tools for real-time interactive video.
AI
Hao AI Lab comments on an independent review of its open-source FastH3 Preview against H3 models.
AI
Four former Google researchers found a public benefit corporation to automate machine learning and science.
AI
Horace He proposes that effective ML techniques increase operational demands on infrastructure teams.
