The main thread is Reflective Practitioner, on Substack. Older academic publications sit underneath. New essays roughly every two weeks. No newsletter cadence theatre.
~ 2,400 words · 9 minute read
Most “agent failures” are not the model losing competence. They are the model losing track of which agent it is supposed to be. Notes on identity drift from a panel system in daily production use, and the verification gate I built to stop it.
~ 2,000 words · 8 minute read
On the difference between a model that has been trained to be agreeable and a colleague who tells you when you are wrong. Why you want the second one. How to ask for it. What it sounds like when you get it.
Forthcoming
Comprehension-based verification mechanism for AI agent context integrity.
Provisional patent application filed with the USPTO on April 28, 2026. Sole inventor. The mechanism is now the verification gate at the boundary of Simpliloquy: if a session cannot demonstrate comprehension of its assembled context, it does not start.
Twelve peer-reviewed papers and reference works in the history and philosophy of science. Most are on counterfactual evolutionary theory between 1880 and 1920: what biology would have looked like had a few intellectual moves landed differently. The full list is available on Google Scholar; selected work below.
Doctoral dissertation · Arizona State University
Studies in History and Philosophy of Biological & Biomedical Sciences
If you read something here and it lands, or it lands wrong, write to me. I read everything.