Research library
Most of what we have built is not deciding anything, and this is the page that says so. Below is the register of every instrument in the engine, grouped by the job it does and labelled with how much weight its output is allowed to carry — then the live board, which is the output of exactly one of them.
Educational and entertainment statistics only. Not investment advice. This is a research model showing its work, for learning and paper trading. It does not know you or your situation, and nothing here is a recommendation to buy or sell.
What the model is built from
The engine is not one program. It is 31 separate instruments, grouped into 8 families by the job they do, and most of them are not deciding anything. This is the register of what exists and how far each part has actually been taken — the maturity of the research, published as it stands rather than summarised after the fact.
A dated snapshot, taken 2026-09-09. Statuses change when the research does, and this page is re-derived by hand rather than wired to a feed.
What the statuses mean
| Status | What it means | Count |
|---|---|---|
| Live | Running in production and feeding something a person or the board actually reads. | 10 |
| Measuring | Records its readings on every run and changes no selection. An instrument here is gathering evidence about itself, and nothing it produces has reached a decision. | 6 |
| Built, not wired | Finished and tested, deliberately not connected. Switching one on is a decision that gets written down, not a setting. | 1 |
| On demand | Run by hand when a question needs an answer. Writes a study rather than a daily artifact. | 12 |
| New | Committed, and has not produced its first output yet. | 1 |
| Paused | Live, and deliberately being held still. The date it stopped and the reason are on the record. | 1 |
The families
Three of these — the resemblance lane, the engine-score lane and the shape sidecars — are three different answers to the same question, built independently and deliberately not merged. Only the first one reaches the board.
| Family | What it is for | Instruments |
|---|---|---|
| Foundations | Shared by every lane below. Nothing here picks anything: they build the day's candidate pool, fetch the price history, and define the readings everything else is written in terms of. | 3 live |
| Resemblance lane | Ranks candidates by how closely they resemble a curated set of historical moments. This is the lane the published board consumes, and the only one that reaches a reader today. | 1 live1 measuring1 built, not wired3 on demand |
| Engine-score lane | Ranks the same candidates by the model's own trend-alignment score, with no resemblance judgement anywhere in it. Built as an independent second opinion; it has never fed the board. | 1 live4 measuring1 on demand |
| Shape sidecars | The youngest idea here: selecting on the shape of a name's trend structure rather than on resemblance or on score. Nine days old, and no forward evidence stands behind it yet. | 1 measuring1 new |
| Hand-given lists | Names a person chose. Each is read with the same live machinery as a model candidate, so the numbers are the same quantity — and each is structurally unable to become forward evidence, because the choice was not the model's. | 2 live1 on demand |
| Renderers | Each reads a stored result and draws it. None of them fetches anything: a renderer that fetched would produce numbers nobody could reproduce from the stored result afterwards. | 3 live3 on demand |
| Board promotion | The single point where a research result becomes a name the product serves. One instrument, one write a day, before the market opens. | 1 paused |
| Probes and studies | Measurements, not evidence. Each answers one question and writes somewhere nothing downstream reads, so a study can never quietly become an input. | 4 on demand |
The board
One card per stock and timeframe, showing where the model stands. Free, no account, running fifteen minutes behind live — measured from the moment the model read the bar, not from when we published it.
The glyph on each card draws its four published readings as one shape — hover it for the numbers. Similar shapes mean similar readings; a missing reading draws as zero rather than assumed healthy; the colour follows the shape, not a good-to-bad scale.
Each card carries two times. Bar closed is the last price the reading saw; Evaluated is when the model read it. The gap is the model working through the board one name at a time — minutes apart is normal.