Tag
Mastra's latest release restores agent state without re-reading the whole conversation. The fix exposes a cost problem most agent users never knew they were paying: every resumed thread re-bills the entire history.
ADK 1.35.0 reads like routine maintenance. The detail worth reading is what it reveals about who sets the pace in Google's agent stack: the model, not the framework.
Vercel's AI SDK now re-validates every redirect hop before downloading a file. The fix is small. What it signals about agent URL handling as a security boundary is not.
On-demand capability loading in Pydantic-AI v1.105.0 is being sold as a performance feature. It's actually an admission that the monolithic-agent pattern doesn't survive contact with real users.
Arize Phoenix v16.0.0 ships Code Evaluators that let users write their own scoring logic in the UI, no deployment required. The real story is what this admits about the state of agent evaluation.
A wave of recent patches across major agent frameworks reveals an unexpected pattern: minor fixes are reshaping ecosystem trust and accelerating modular adoption.
Hermes Agent's latest 'Tenacity Release' shows that the path to more durable agents lies not in preventing failures, but in accepting them as inevitable and building around their reality.
