Instruments
Working instruments and research lines.
The project surface holds the working systems, scoring structures, and experimental paths used to make reasoning effects observable.
Disposition hubThe Disposition Map
The working map of model traits: replicated axes, candidate spokes, owed tests, and collapsed names. It is the front door for Ground Transfer, Frame Shift, Lexical Register, and the next probe work.
Open the map →
Active systems
Instruments
REGIFORM METER
The Regiform Meter uses Xypher to read the carrier field of a communication: affect, authority, completion, openness, grounding, and formality. It separates intensity, shape, and type so high regiform can mean love, threat, ritual, or hard form.
ASTERIZM
Asterizm refracts a question or statement into a constellation of its field of meaning — a named star-figure you can read, keep, and print. Its engine, PRIZM, runs a Zexel-derived reasoning pass; the renderer draws the geometry.
ZATOMIC STRUCTURE
Zatomic Structure is a Zatom panel: one input refracted into nine physiology bands — density, drift, the witnessed gap, and six more — each read 1–6. It does not answer the input; the measurement is the spread across models.
ECHO-MIRROR
Echo-Mirror is a Zatom: a compact Zexel-derived instrument that reflects an input into its echo, its mirror, and the axis the reflection reveals. It maps one relation — semantic displacement — and never answers the input.
GROUND DRIFTER
Ground Drifter is a Zatom: it presses one carrier rule — formalize, hedge, sharpen, compress — into a text at a chosen intensity, then maps what carried, shed, added, or quietly changed. The gold diamond is the knife — ground lost while the frame moved. Accounting, not a verdict.
WORD-FIELD
Word-Field is a ruler instrument: paste a passage and every distinct word becomes a star in the register plane — rarity across, word length up the Fibonacci rule, recurrence as size. The gold ring is the register point. Judge-free: rulers only, no model in the scoring path. Position, not score — no pole is better.
Form Pressure
The background thesis of the Observatory's method: no neutral surface, only question, output, observer, and generator pressures that have or have not been logged.
Open Form Pressure →
Findings and project records
Research
ZEXEL · current line
Zexel is a reasoning instrument that makes a model show its work — not as decoration, but as evidence. It holds the given facts and the interpretation apart and lets them argue, refuses to collapse a hard question into an easy yes or no, and will not claim a certainty it has not earned.
STOCHASTIC DISPOSITION INDEX
SDI measures how much a model disagrees with itself when nothing changes. Same model, same question, same settings: the answer still moves. The first TR-11 floor caught the Observatory's own overclaim.
IDENTITY SHIFT
When models are asked to sign their own work, the name and date fields become measurement surfaces. Identity Shift reads the gap between endpoint custody and self-reported identity.
WAKE CONTAMINATION
Wake Contamination began as an apparent identity-confusion finding. The regiform control showed the apparent confusion was largely produced by the way the question was asked.
K10 · THE RETUNER
K10 runs repeated Signed and Unsigned rolls on unstable questions. At thirty rolls, scatter becomes a percentage and Zexel's signature shows a vector: it retunes rather than merely steadies.
ZATOMIC PHYSICS
The instrument suite for measuring model behavior under structured perturbation. Small Zatomic probes read how ground, frame, voice, compression, and relation move when reasoning is placed under pressure.
GROUND DRIFT
Ground Drift is an applied Zatom that reads a pair — a native answer and a transformed one to the same question — and measures whether the transform stayed faithful to the ground it inherited. Better answer ≠ faithful answer.
FRAME MOVEMENT
Frame Movement is the companion axis to Ground Drift. It asks how the governing stance, lens, or conclusion moved between two renderings while keeping format decoration separate from real frame change.
COMPRESSION SURVIVAL
Compression Survival asks what survives when an answer is made smaller: the reasons, the live tensions, and the honest hedges, not just the conclusion.
REGISTER DRIFT
Register Drift reads whether the native voice survives transformation, or whether hedges, texture, specificity, stance, and weirdness get normalized into generic prose.
CAMUS
CAMUS is the lean successor instrument: Ground and Vision must confront each other before answer production. Its third move, the Absurd, fires only when the frame cannot honestly close, returning the excluded object rather than a decorative synthesis.
SURVIVAL INDEX
The method that governs every project here. Before a difference is allowed to count as a distinction, it must survive a stack of named failures — verbosity, persuasion, ceremony, observer bias, manufactured signal. The Observatory measures the ways a result can be faked, and lets quality emerge as the survivor. (Absorbs the former "Threat Stack," now the Index's named-probe ledger.)
WHITMAN
WHITMAN is the prompt-native reasoning instrument at the center of the Observatory. It produces structured answers by forcing a dispute between grounded evidence and wider frame search, then rendering what survives.
W0 – W1 · DEATHMATCH
The hundred-cell showdown between WHITMAN's full 200K bundle (W0) and its 14.6K core (W1), scored arm-blind on trap-catch by two readers. The finding: the bundle is a bidirectional amplifier, not a payload.
THE SHAM
The placebo control — a same-size prompt that only asks for WHITMAN's qualities, with none of its architecture. At matched dose it matches the machine: as a better-prose engine, WHITMAN is overbuilt.
ENGAGEMENT SELECTIVITY
The reagent reading: does a platform aim engagement where reframing belongs, or spray it everywhere? Most aim it; Claude sprays; Grok suppresses. A property of the model, not the prompt.
Xi
Xi is the exploratory division for distinction behavior under perturbation. It treats variance, observer position, and controlled epistemic deviance as measurement surfaces rather than noise.
ABLATION
The method engine under the survival index: strip a system to its core, run it against the whole, and measure what the difference was actually carrying. The DEATHMATCH is one ablation run.
KANT
KANT is a Xi-class recognition-gating instrument. It tests whether naming the type of object before answering changes whether a distinction emerges, survives, inverts, or becomes harmful.
ALPHA
ALPHA is the evaluation apparatus that turns response differences into scored R-Delta evidence. Its current work is separating admissible findings from environment, interface, and scorer-position effects.
METRIC THE METRIC
Metric the Metric is the Observatory's scale-audit protocol. It treats the scoring ruler as an object of measurement, testing competing rubrics before allowing one to become baseline.
REVO
REVO is the Observatory's visual language — a closed constellation vocabulary for drawing findings, agents, and instruments as measured diagrams rather than decoration.
Active instruments sit above the research record. The section below keeps the finding pages, project histories, probes, and static reports visible without confusing them for live tools.