A reasoning engine, not a chatbot. Built for the questions where surface answers fail.
Medical mysteries, institutional behavior, environmental signals. Boaz investigates questions that require more than retrieval -- they require reasoning about what is missing, what is contradictory, and what would change if it were wrong.
Traces the cause
Follows evidence chains from symptom to source. Every link is cited, every inference is flagged.
Checks for distortion
Looks for the ways evidence can mislead: selection bias, confounders, missing data, motivated reasoning.
Maps what is missing
Identifies the evidence that does not exist yet -- and tells you what would change the answer if it did.
Six steps. Each one is a deterministic operation you can replay.
Why does my joint pain return, almost without fail, on Tuesdays?
Move whatever you do on Mondays by two days. If the pain shifts to Thursday, it is candidate #1.
What we will always do
- Plain language first. Jargon is a last resort.
- Show the work. Every claim has a way to check it.
- Short sentences for the scary parts. Long sentences for the careful parts.
- Name the limit before someone asks.
- Never sell. State, and let the work pull.
What we believe, out loud
- We believe environmental exposures are under-studied.
- We believe patient testimony is evidence, not anecdote.
- We believe institutional inertia causes real harm.
- We believe falsifiability is the price of admission.
If we cannot show the path to a sentence, we do not write the sentence.
Independent researcher working at the intersection of AI philosophy, medical epistemology, and computational reasoning. Published in neuro-symbolic AI and phenomenological methods for contested diagnoses.
The work so far
Published research, datasets, and manuscripts in preparation. Every claim is sourced. Every method is documented.
Join the beta.
Run a question through Boaz.
Limited cohort. Bring a real question -- medical, institutional, environmental, historical -- and a willingness to read the trace.