All work
Systems that spend
What a 2 a.m. war room teaches about governing AI.
Before AI agents, our war rooms were smart-contract pools holding $3M. The discipline is identical: systems that move value need boundaries, drills, and a second pair of senior eyes.
$3M
value under review scope
$900K
at-risk value caught before mainnet
1
checklist that became a practice
Problem
Timeswap Labs, our first major client, needed review of contracts guarding $3M in liquidity. A novel AMM design with real funds approaching and a roadmap that could not slow down.
Approach
- 01Mapped who could move funds and under what conditions — the same trust-boundary exercise we now run for agent permissions.
- 02Built adversarial scenarios instead of reviewing the happy path twice.
- 03Caught a testnet flaw putting roughly $900K at risk, then paired on the fix.
- 04Left a repeatable review checklist the team still runs — the ancestor of our agent eval harnesses.
“A 2 a.m. debugging session that saved the pool. That is the engagement in one sentence.”