The math, live.

One currency, one fee point, every inference model behind it. The treasury earns 0.665% of every $HERO trade. Pick a target and drag the slider to see how fast trading volume funds it. $10M proves the mechanism. $100M is a real frontier run, the kind a new lab needs to raise to attempt. $1B is an endowment that funds open training forever. These are scenarios, not forecasts.

Live from DexScreener (Base)

Runs fund mostly through the fee engine above. Product margin on each inference call and the treasury's own $HERO holdings add to it, and every run logs a prompt-and-output pair as opt-in synthetic training data. The fee point is fixed, so the only variable is volume: the $10M first run funds inside a year at sustained $5M–$50M/day, normal for a memecoin that catches attention. The $100M and $1B targets need the token to trade like a real large cap, which is exactly what the benchmarks below show.

Scenarios use real fee mechanics and live market data. They are illustrations of the model, not predictions of price or volume. Not financial advice.