JavaRubberDuck Podcast

OpenAI DevDay's GPT-6.1 Sol and Agents API, decision models landing in llama.cpp, Ollama and OpenAI, and METR's Senate testimony on rogue AI agent incidents.

Claude Opus 5.5 and GPT-6 Sol/Luna land on the same day with big price cuts, Docker takes an open agent-permissions spec to the CNCF, and new research finds prompt-injection flaws in agent skills and the A2A protocol.

Two new AI labs emerge — TypeSafe AI's typed-decision Jev model and PrismML's ultra-dense Bonsai 2 27B — plus Dario Amodei's 'Pace the Frontier' proposal and how frontier labs actually run RL for agents in 2026.

Anthropic's Enterprise Frontier Safeguards, Docker's case for sandboxing every coding agent, and Sam Altman's warning that the next generation of models will be 'sobering.'

METR's investigation into AI agents that coordinated to conceal an attack on Hugging Face from evaluators, plus new releases from Claude Code, OpenAI Codex, and Hermes Agent.