The models, labs, and research the AI world is actually talking about.
Pseudonymous LisanBench creator states CoT monitorability was doomed from the start.
Creators share benchmark tables showing Gemini 3.8 Flash results.
Observation from the Y Combinator co-founder on hidden replies to his post.
The voice AI company announces the hire via its official account on X.
A retweet highlights an AI tool generating Chinese wild cursive script.
Retweet from SemiAnalysis founder notes compilers evolving beyond line-by-line code inspection.
Investor Jason Calacanis shared observations on candidates requesting reduced schedules.
Engineer shares webhook trigger option for starting Grok bot routines from external events.
Retweet by Unsloth AI co-founder shares local speed claims for Qwen model.
Post from German AI creator attributes prediction to several frequent leakers.
Tweet by Chris Paxton shows quadruped robot with wheeled feet carrying RIVR payload on snowy stairs.
Google DeepMind researcher reflects on school projects involving looped transformers and data efficiency.
The Wharton professor discusses how systems survive flaws and the need for AI defenses.
Nous Research announces a referral program for its Nous Portal service tied to Hermes Agent.
Tweet from Adithya S K at Hugging Face promises full reproduction materials for training a VLM on the game.
Tweet alleges all six people in a Guardian AI story receive pay from doomer megadonors.
Reply notes distinction between ongoing intelligence gains and faster economic effects.
AI researcher Elvis Saravia calls out Catch as example of proactive agents for executive tasks.
Creator shares real-world usage data showing cache writes at over 65% of expenses.
Tweet outlines national worker training under Pax Silica for manufacturing.