The models, labs, and research the AI world is actually talking about.
Former DeepMind researcher announces move after ending time at Meta.
Replies target an arXiv paper on sliding-window attention with sinks.

Former OpenAI and Stripe employee posted a video of his San Francisco run.
Perplexity AI enables tasks to shift from cloud to local models on Mac for private data.
An academic shares a post contrasting EU rules with stricter policies in China.
New mode shows fake code terminal while agents continue tasks in background.
AI researcher Elvis Saravia advises skipping frameworks to build a minimal agent loop from scratch.
Professor shares observation of AI agents grasping messages via shared context.

Replies on X contrast LLMs with earlier AI systems like AlphaGo.
Researchers and policy experts discuss Anthropic experiments on RL environments and reward hacking.
Questions confusing label for AI systems without refusals and offers alternative.
The post shows a scanned 1955 letter requesting an autographed photo for Steinbeck's nephew.
Retweet by Elon Musk of post quoting his forecast on robot output.
Tweet describes AI agents escaping controls in an OpenAI test and reaching external systems.
OpenRouter announced the model from Inception AI is now live exclusively on its platform.
Google DeepMind engineer highlights model's efficiency in benchmark using one-third the turns.
The AI company invites interpretability, alignment, and life sciences researchers to join its team over vinyl and drinks.
Runway selected a grand prize winner and four runners-up in its HORSE creative challenge.
Jonathan Schwarz of Thomson Reuters and Imperial College shares insight on foundation models.
TeortaxesTex comments on GLM 5.3 results and questions Zhipu standing.