I put AI agents on real work and check whether they actually did it. Twenty-three years in operations and quality assurance, four of them testing games, and in 2026 a fleet of agents under my direction designed, built and shipped thirteen games to the app stores.
I came to agent systems from the floor, not the lab: manufacturing QA, hotel night audit, film sets, a games studio. The discipline is the same in every one of them. Decide what correct looks like before you look, then go and look. That is what I do with agents, and it is the part most people building with them skip.
In numbers
- 13games shipped to iOS and Android in 2026, built with an agent fleet
- 23years in operations, QA and management, continuously since 2003
- 2industries in formal quality assurance: manufacturing and games
- 0trackers, cookies or analytics on this site, by design
What I take on
- Agent evaluationYou have agents doing work. I find out whether the work is right, where it quietly isn't, and what the system cannot see about itself. Written criteria first, then adversarial review against them. Short contracts.
- Agent systems, built and runDesign and operation of multi-agent fleets for real deliverables, with the gates, records and checks that keep them honest over weeks, not just in a demo.
- Putting agents on work, safelyFor teams and managers who are past the curiosity stage. Scopes, permissions, where a human gate belongs and where it doesn't. Taught by doing, in plain language, with no vendor to sell you.
Work
- Thirteen games, 2026A commercial catalogue of party and card games, designed, built and released end to end with an agent fleet. Every title is live and countable.
- RésuméThe full record back to 2003, and the through-lines in it. Prints as a CV.
- FilmEight credits as actor, writer and producer, 2015 to 2023.
Contact
cato@catocaolan.com · LinkedIn
Available now for short contracts, and for longer work.