Multiplicity
Writing, systems, and executive AI.
A site for long-form writing on behavior analysis, AI, and operations; for the software and coordination systems behind that work; and for an executive intensive on implementing AI in real operating contexts.
Recent writing
Two papers on how AI actually behaves, and what building with it takes.
A machine that says sorry
Anthropomorphic AI gives us familiar but flawed controls. The current risk is less that users believe the machine is sentient and more that the interface cues the wrong repertoire. Apologize, reprimand, or scold — each one changes the next context the model transforms, not a communication partner's behavior.
The field that taught machines to learn
Behavior analysis, artificial intelligence, and the cost of not building. Reinforcement learning traces back to Thorndike and Skinner. The people building modern AI know this. Behavior analysis has not made that lineage central to its own account of AI — and outsiders are building the measurement layer around autism care without it.
Projects and systems
Software, workflows, and applied systems
Current work includes ClawSuite Relay, a coordination layer for multi-agent systems. Older projects and rougher experiments are preserved here too.
Executive AI
An intensive for operators with real responsibility
A short, hands-on intensive for owners, founders, and senior operators who want to move from vague AI exposure to a working implementation path.
Lab — in planning
An AI benchmark for behavior-analytic reasoning
General AI benchmarks measure breadth. They do not tell us whether a model can distinguish behavioral measurement systems, interpret ABC data, or notice subtle category errors in clinical prose. Early-stage research program toward a publishable benchmark.