Methodology.
How an answer becomes a finding — and what the instrument refuses to do.
The claim under test
The lab's thesis is that the worst version of a person is a default mode: under pressure, deliberation goes offline and the most-practised behaviour runs. If that is true, behaviour under pressure is predictable — and therefore measurable, comparable, and mappable.
The instrument
Conversations are adaptive conversations, not forms. Each answer is reflected back and routes the next question, so the path flexes — but every branch scores on the same eight-pattern engine. Paths differ; the mathematics does not. That is what makes a Work conversation in Lagos comparable to a Love conversation in Las Vegas.
What gets recorded
Pattern, depth score, all eight pattern scores, which conversation, and a coarse country. Nothing else. No names on results; typed answers are studied as language in aggregate, never published attached to a person. Consent is stated at the door, in plain language, because a lab that studies honesty has to practise it.
The triple rule
No assessment ships unless it pays three ways at once: a real verdict for the person taking it, structured data for the lab, and comparability for the science. A question that flatters the user, bores the dataset, or breaks comparability is cut — however clever it is.
The model register
Each conversation is a model of one room of a life — named, the way instruments are named, and versioned like one. The family is stone: the Monolith's own material, cut for each room. The register is public, so any result can be traced to the exact instrument generation that produced it:
- The Mirror — Obsidian 1.0 · the flagship; the master pattern.
- Work — Slate 1.0 · Money — Marble 1.0 · Love — Opal 1.0 · The 2AM Conversation — Moonstone 1.0
- Family — Bedrock 1.0 · Your Circle — Hearth 1.0 · The Body — Granite 1.0 · Ambition — Flint 1.0 · The Past — Amber 1.0
- The Peak — Topaz 1.0 · Best Mode (TBVY): the mirror instrument. It scores six peak drivers instead of the eight patterns, and files a prescription instead of a verdict.
A minor release (1.0 → 1.1) tunes questions and branching against live data. A major release (1.0 → 2.0) changes the engine or the depth calibration — and is published as a finding, with what improved and by how much. Models never change silently: a result stamped 1.0 must mean the same thing in ten years as it does today, or the Index is fiction.
Limits, stated
Self-report is a lens, not a window. The lab's findings describe reported behaviour under remembered pressure, and the published work says so. The panel and longitudinal studies on the roadmap exist precisely to close that gap over time.