Methodology
The real product is a research pipeline. Hearing loss is where we point it first.
MISHA is a bet that a disciplined, AI-augmented research pipeline can compress the early, expensive part of science. We are testing that pipeline in the open, on problems where the answer is checkable, and applying it to my son's genetic hearing loss. The method is the point. The disease is the proof.
How it started
The first real test was STRC. I ran the computational analysis on my son's gene the way I would run any research problem, and then I did the thing you are supposed to be too intimidated to do: I sent it, unannounced, to one of the most respected hearing-genetics labs at Harvard, and to leading researchers in Shanghai.
The responses came back serious and engaged, from people who had every reason to ignore an unknown parent with a spreadsheet. That was the moment the idea flipped. It was not a fluke, and it was not a hobby. The pipeline produced work that credible scientists on two continents took seriously. If it could do that, it could do it across rare disease, for families who will never cold-email Harvard themselves.
Computation is a lever, not an oracle
The discipline underneath all of it: the model is confident even when it is wrong. Teaching the method at HKU, I watched an AI generate a clean, plausible formula that was simply false, caught only because the real measured data existed to check it against. So the pipeline is built around verification, not trust. Books and ground truth first, computation second. That single rule is what separates research from generated confidence, and it is the reason a busy lab can trust what we send them. When anyone can generate a plausible hypothesis, the scarce thing is one that has already been checked. We hand a researcher the survivors, not the noise.
Why we prove it on mathematics first
Math is the honest proving ground, because you cannot argue with a proof. We point the pipeline at famous long-open problems, the kind Erdős posed, and we formalize every step in Lean 4 so a computer's proof kernel checks it line by line. You cannot bluff a kernel.
The honest status, stated the way we state everything: no famous problem is solved. What exists is real partial progress, finite cases and machine-verified certificates, and a public log of what did not work. One line of the work has reached a preprint. We publish the attacks and the dead ends, because a pipeline you can only see when it wins is a pipeline you cannot trust.
The same pipeline, pointed at hearing genetics
On STRC and adjacent hearing-loss genes, the pipeline does the same three things it does on a math problem: rank the hypotheses worth testing, predict the structures with AlphaFold3-class models, and find the druggable pockets, all before anyone spends wet-lab money. The outputs are public. This is the engine inside the Research Lab program.
Why this matters for a foundation
A one-person foundation cannot ask you to trust its polish. It can show you a method that works on checkable problems, run it in the open, and let the work speak. That is the whole wager: build the pipeline once, prove it honestly, and point it at the diseases the market leaves behind.