
Anthropic Trains Claude to Reason Like a Research Chemist
New capability work demonstrates how large language models can be specialized for scientific research domains, raising questions about AI's role in accelerating discovery.
Launches, releases, and hands-on coverage of the AI tools you'll actually use, coding assistants, agents, creative tools, and the infrastructure underneath.

New capability work demonstrates how large language models can be specialized for scientific research domains, raising questions about AI's role in accelerating discovery.

The AI company has suspended access to two advanced models following a federal directive, raising questions about government oversight of frontier AI development.

Cybersecurity findings about AI vulnerability to prompt injection trigger government export controls on advanced language models.

An experiment shows AI can build functional software in minutes, yet struggles with basic debugging without user intervention.

Vulnerabilities, patches, and a hardening checklist for Python teams deploying AI agent backends.

Export controls and reported safety vulnerabilities force sudden shutdown of Fable 5 and Mythos 5, raising questions about AI deployment oversight.

New benchmarking tool aims to streamline how researchers test and iterate on language models throughout their development cycle.

Engineers demonstrate how kernel fusion and compiler-level optimizations can significantly reduce computational overhead in deep learning models.

A new job scheduling system lets ML teams run training and inference workloads directly on Hugging Face infrastructure, reducing vendor lock-in.

The startup's compact coding model aims to bring specialized language capabilities to programmers building with limited resources.

Testing exposes critical gaps in how speech recognition handles code-switching, a challenge for global voice assistant deployment.

Researchers demonstrate how autonomous agents can coordinate between separate AI services to generate complex visual outputs without human intervention.

New building-block approach lets developers compose custom AI agents from reusable components without starting from scratch.

The AI leader joins Anthropic in pursuing IPO registration, escalating competition between the industry's two most prominent startups.

A new collaborative framework aims to standardize how AI agents learn through interaction, gaining momentum across the developer ecosystem.

Researchers explore what happens when language models interact within shared digital marketplaces, revealing surprising patterns of coordination and control.

A shooting survivor's legal action reveals critical gaps in automated gun detection systems deployed to protect students.

Chinese AI startup's newest release demonstrates measurable gains in precision benchmarks, intensifying competition in the large language model market.