A closed scientific programme on what about Bitcoin’s future is actually predictable: code, evidence, failures and the final adjudication of 24 predictability dimensions.

It asks a better question than “will price go up?” — given the state now, what is the probability price touches an upper barrier first, a lower barrier first, or neither, within 2 hours?

THE ONE CONFIRMED RESULT
On a sealed, outcome-blind holdout of 19,672 decisions the models never saw during design, the frozen model gained 0.021827823 bits per decision, 95% CI [0.012560632, 0.031356671], with a better Brier score — under gates fixed before the data was acquired. Baseline entropy is about 0.862 bits: it removes about 2.5% of the remaining uncertainty. Small, real, honestly measured.

WHAT FAILED
Terminal direction: rejected. A billion-trade cross-exchange tape: +0.001123 bits at one minute, gone by five. Aggregate derivatives: subtracts information at every horizon. One campaign was destroyed by the project’s own audit, its numbers withdrawn.

INSIDE
23 research scripts, 184 JSON artifacts, 7 notebooks, 19 pre-registered mission briefs, 133 linked pages in an Obsidian-ready vault, English translations of the Spanish canon, and an offline demo running in a second.

NOT a bot, a signal service or a profitable strategy. No PnL, cost, slippage or capacity figure exists anywhere in it. Market data is not included. The project is closed: no updates.

NOT FOR anyone looking for signals, entries or copy-trading.