doorss

Anthropic publishes an initial update on Project Glasswing, their research initiative. The post details early findings and directions from the project.

Andrej Karpathy announces he has joined Anthropic, marking a significant move for the prominent AI researcher.

Microsoft acknowledges that AI agent costs currently exceed what it would pay human workers for equivalent tasks, raising questions about the economics of AI deployment.

A research paper examines how LLM coding agents progressively lose adherence to constraints when generating backend code, revealing fundamental fragility in agentic code generation.

Intuit plans to eliminate over 3,000 positions as part of a restructuring to prioritize AI development across its products.

George Hotz writes about the permanent degradation of online content quality due to AI-generated slop, drawing a parallel to Usenet's 'Eternal September.'

An analysis of how LLM-generated text exhibits distinctive patterns including overuse of em dashes, serving as a stylistic fingerprint of AI writing.

Armin Ronacher describes the frustration of receiving AI-reworded bug reports from open source users, where LLMs obscure the actual observed problem with verbose, unhelpful reformulations.

U.S. Customs and Border Protection publishes its updated directive governing how agents may search electronic devices at the border.

The FTC settles charges against Cox Media Group for deceiving customers about an AI marketing service that claimed to use 'active listening' through device microphones.

Metan tekoälylaseilla on kuvattu naisia salaa. Huolet lasien väärinkäytöstä kasvavat, ja myös Metan työntekijöille päätyy arkaluontoista materiaalia.

Helsingin Kalasatamassa syttynyt kerrostalopalo tuhosi 62 asuntoa ja koko rakennuksen. Poliisin mukaan palo sai alkunsa kaasugrillistä parvekkeella.

Suomi joutuu maksamaan lainoistaan riskilisää, jonka hinta on satoja miljoonia euroja. Syinä ovat Itämeren turvallisuusriskit, valtion velkataakka ja kilpailu Saksan kanssa.

Helsinki aikoo kieltää fossiilisten polttoaineiden mainonnan kaupungin kilpailuttamissa mainospaikoissa, vaikka se vähentäisi mainostuloja.

Google announces new AI-powered advertising formats in Search results and expands its Direct Offers pilot program at its Marketing Live event.

Bluesky's engineering team details how they build and serve the 'For You' algorithmic feed, explaining the AT Protocol's open feed architecture that allows anyone to run custom feed generators.

The Anthropic Institute shares its research agenda, focused on investigating AI's real-world economic and societal impact using data from within a frontier AI lab.

Anthropic donates Petri, their open-source alignment testing toolbox, making it available for the broader AI safety community to test any large language model.