Book and research direction
What Agents Want
A developing inquiry into artificial agency and human–AI relations.
Subject
Agents & Judgment
Making AI-assisted decisions inspectable, and asking what agents want.
What can an artificial agent be said to want, and who remains responsible when a model gives advice? These projects treat agency, evidence, authority, and judgment as related questions that should not be collapsed into one another.
The work ranges from a developing book idea to decision records and bounded experiments. A model output is evidence to examine; it is not authority by itself.
Writing
Projects
Making AI-assisted decisions inspectable: what was known, which model was consulted, who judged the result.
Private beta · limited access
Plugin Studio
A guided builder for turning an idea into a personal ChatGPT plugin. Describe what it should do, shape its instructions, and try the draft experience.
Concept
MyMonkey
A concept for small agents that help with repetitive routines after watching how they are done.
Concept
AI Mirror
Seeing, and correcting, what an AI appears to have inferred about you.
Method kit, under repair
AI-IA Experiment 001
Does comparing personalised and unpersonalised AI answers help people judge advice?
Scripted demonstration
Decision Adapter
Bounded model advice with a record of each call. It advises; it does not grant authority.
Prototype
Ontogony Diagrams
Deterministic, editable diagrams from structured graphs, and a source-aware system explorer.
Prototype
Resolution Observatory
Investigate evidence, apply scenarios, and replay an investigation offline.
Internal
Owner Brief & Constellation
A read-only overview of changes and review needs across projects.
Archive
Canonizer
Reconstructing decision state from competing claims, shown with a fictional bank.
An internal harness for keeping agent-assisted development efficient and under control.
Compares AI forecasts and virtual allocations with market data. No real money; no performance claims.