Skip to content
MustHave.ai
MustHave Hoodie Seal
THE MUSTHAVE NEWSROOM

AI news

What’s new. What matters. What do you think?

Top stories

By member votes

PAST-Bench makes AI agent memory prove it helped

PAST-Bench tests agent memory across clean sessions. A new TEPA update shows why useful memory also needs explicit validity, revocation, and an audit trail.

Gemini Managed Agent hooks fail open. Build the boundary outside them

Gemini Managed Agent hooks can block sandbox tool calls, but failures and timeouts allow the action. Here is the safer control design.

ChatGPT is moving from asking to doing. The country data shows where

OpenAI's first country-level ChatGPT usage data shows work shifting from asking to doing, multimedia rising, and older users gaining share.

GitHub Copilot Metrics: Separate Named Tools, Client Requests and Review Outcomes

Copilot metrics now identify more custom tools and break out usage by client. Check telemetry coverage, historical gaps and request definitions before comparing teams.

LangChain Managed Deep Agents is public beta. Shared memory needs a policy

LangChain's hosted Deep Agents runtime is now public beta. Here is what it manages, and the memory, identity, tool, trace, and schedule policies teams still own.

California AI transparency law is live. Here is who must comply now

California's AI transparency duties became operative August 2. Here is the provider threshold, disclosure system, platform calendar, penalties, and builder checklist.

Claude Code cross-session messaging is live. Treat the inbox as a permission boundary

Claude Code sessions can now exchange plain-text messages. Here is how delivery, permissions, remote replies, held messages, and administrative controls work.