Skin in the Game Made the AI Supervisor Worse
A controlled experiment on the cost of wrongful blocks and the incentives that shape an AI reviewer.
Oversight & accountability
Ideas across time · About
About
Born in the Netherlands, at home in Chicago, with a lifelong habit of asking one more question.
A life between worlds.
I’m Dinand Tinholt, an AI consultant, researcher, writer and builder. Originally from the Netherlands, I now call Chicago home.
My work at Capgemini brings me into conversations about how organizations use AI and data, and how those choices affect the people doing the work. I’m especially interested in decision intelligence and the relationship between human judgment and increasingly capable systems.
Beyond technology, my interests lead me into history, literature and philosophy. Old books and maps share shelf space with ideas about the future. I find inspiration in art, from Warhol’s repetitions to the questions Banksy leaves on a wall.
This is a place to bring those interests together, through essays, research, things I build and notes from what I read.
All writings and statements on this site, and in my publications elsewhere, are my own. They are not attributable to Capgemini.
A body of work
Five years of thinking in public.
Writing began with data culture in 2022 and has moved, year by year, toward agents, oversight and simulation.
- 115
- Published articles
- 16
- Research studies
- 7
- Systems & instruments
- 3
- Recent public appearances
Articles per year
Counts include only articles with a known year.
What the writing returns to
The Delegation Cliff
90,880 simulated designs of an AI-delegating organization. At a 95% quality floor they degrade gradually; at 99% the same design space develops a cliff.
Oversight & accountability
AI Agents May Have a Reason to Trust Each Other
On a paper led by Alexander Meulemans: why foundation-model agents may cooperate in a one-shot Prisoner’s Dilemma, and what that means for large populations of agents.
The agentic enterprise
Nobody Slows Down First
Six days apart, the same industry asked Washington to keep AI capability spreading and to build the machinery for holding it back.
Models & the AI market