Back home
Notes
Notes
Short, single-idea takes — the things I keep circling, set down quickly. For the longer pieces, see the writing.
A human in the loop has to be able to say no
Otherwise it isn’t oversight. It’s decoration.
June 20262 minHuman-in-the-loop
Most AI risk is just access you forgot to scope
The scary failures aren’t clever. They’re an agent doing an ordinary thing with permissions nobody meant to grant.
June 20262 minSecurityAgents
An audit trail you can edit isn’t an audit trail
If a record can be quietly changed after the fact, it was never really a record.
June 20262 minAuditGovernance
Pilot AI where being wrong is cheap
Learn the failure modes on work that doesn’t hurt when it breaks.
June 20262 minAdoption
Sovereignty is just accountability you can verify
A strategy you can’t inspect is a wish. The verification is the whole thing.
June 20262 minSovereign AI
A demo should be honest about what’s real
The engine can be real and the inputs staged — as long as you say which is which.
June 20262 minBuildingHonesty