A map of what this platform can detect, not a claim that any of it works. Nothing here is validated. A library this size is a multiple-comparisons machine — at p<0.05 roughly one sequence in twenty looks excellent by chance, and this platform has already read +0.323R in sample and −0.048R out on a 900-config search (ADR-069). Each atom and each sequence carries what has actually been measured about it; an empty measurement means unmeasured. The candles on each card are synthetic — a generated picture of the sequence's own stages, drawn so sweep, then structure breaks, then the gap is retested is legible at a glance. Hover a numbered badge for that stage's role.

Every observation the hybrid engine can extract, in every state. This describes the platform, not the market in the toolbar — it is the same list on every symbol and it does not reload when you inspect one. An atom is one stamped reading taken from a layer that already computed it; an ordered sequence of them is a pattern, and those live in the Patterns tab. Each row carries what has actually been measured about that reading, and an empty measurement means unmeasured, which is the default. Nothing here is a signal. Uses counts sequences that name the atom, not ones that need it: a stage naming three alternatives counts for all three, so choch_break reads 78 and no sequence depends on it. The candles beside each row are synthetic — the same generated picture the Patterns tab draws, for one atom instead of a whole sequence, because a sweep and a break differ only in whether price held beyond the level it took and that is a picture rather than a sentence. Direction is a drawing convention: atoms publish a direction per event and the registry declares none. Hover a diagram for what it does and does not claim.