Mark Fussell
Founder, Diagrid at Amplify Partners
Diagrid @diagridio
#Observability tells you what your agent did. It can't prove the record wasn't altered, and it can't govern what the agent was allowed to do in the first place. For high-risk AI under the EU AI Act, proof and control are exactly the layer that matters, and it's the layer Diagrid
Jul 16, 2026 Β· β₯ 7 Β· β 1
Diagrid @diagridio
A LangGraph agent fails partway through a run. The checkpoint tells you where it was. Getting it running again is still your job to build. @mfussell and @yaronschneider are running a live demo: break a LangGraph agent mid-run, add Diagrid, watch it resume from the exact step
Jul 15, 2026 Β· β₯ 2 Β· β 1
Diagrid @diagridio
Your agents are in production, invoking tools on one shared key that opens every door and records nothing. So when one goes off-script at 2am, you can't say which agent did what, or under whose authority. Jani Janakiram, MSV, on why security leaders belong in the agent
Jul 14, 2026