What's strictly AI: models and their updates, AI-specific tutorials, running models locally, agents, and model evaluation. Always with a test and an example.
An agent can look perfect in the demo and still blow past cost limits, get stuck in a loop, and fail silently in production. Google details, in a playbook published on October 15, how to structure a real evaluation beyond manual testing.
Liquid AI released a 24-billion-parameter model that activates only 2 billion per token, designed for agents that call tools and run continuously on local hardware.
The June 18, 2025 revision of the Model Context Protocol specification turns every remote HTTP server into an OAuth 2.1 resource server. Anyone still validating a static token in an environment variable has a concrete migration path, as well as cases where this doesn't apply.
End of the 2 most recent news stories. Subscribe to the RSS feed to follow along in real time.
