Finding data
RTM Lab reconstructs an information state as an architecture, not a collection of matches. Enter records, times, events and the observer registry as documented; declare roles and representations; the engine searches only lawful paths, preserves dependencies and chronology, and exposes convergence, loops, re-entry and reconstruction under the locked Search Contract.
data type + number or Hebrew text · automatic gematria
Only for text-span / nesting analysis — not for the datum text above
Evidence — structured results
RTM Evidence is a relation inside a reproducible architecture. Evidential force increases when distinct sources, representations and roles converge on the same event/observer, when nodes recur through non-identical paths, or when later information reconstructs a component of an earlier state. The network view exposes structure; it does not create extra Evidence merely by containing many edges.
The main card shows the complete equation first. Equivalent algebraic forms are shown when they make the closure easier to read. Long explanations stay collapsed.
Find stays independent. Then run 06 · Statistics; only then does RTM AI unlock.
Strict structural shortlist
This presentation-only layer promotes topology-rich findings: representations, networks, loops, convergence, symmetry and repeated residual convergence. Ordinary arithmetic cannot qualify by itself.
Automatic Full Reconstruction
Runs directly from the frozen inputs. No reference Manifest is required. Structurally equivalent algebraic orientations are merged into one relation.
Run summary
Discovery — broader search space
Discovery is a broader candidate-search space. It is not automatic RTM Evidence: candidates must satisfy the Search Contract, semantic roles, dependencies and relevant controls before statistical interpretation.
Evidence Network
Use the Graph button on a closure to highlight its path.
External reference QA
Optional researcher tool only. Normal analysis and Full Reconstruction do not require a Manifest.
JSON · external fixture · not embedded in engine
Rarity audit — five separate answers, without mixing them
The Audit asks whether the observed architecture survives observer replacement, role changes, event-space alternatives and equivalent search freedom. Every Null world receives the same search contract and path dependencies are preserved. The purpose is not merely to obtain a small p-value, but to test whether the same role-preserving topology is difficult to reconstruct in alternative worlds.
How extreme is the finding on one fixed multivariate rarity score? Each feature receives an upper-tail rank in the same Observed+Null pool, the ranks are combined by one fixed rule, and the primary p is the symmetric conformal rank of that combined score.
Can a Null world reproduce the same three-closure geometry anywhere: two sibling R24 bridges plus an R22 return to their common bridge?
Once you declare what was frozen and what could vary, the engine evaluates that exact event space — exhaustively when feasible — and requires the same slot-bound network.
Each random world may keep the strongest topology it discovers. This prevents selecting only the most impressive observed closure after the fact.
Separately, the audit checks the same topology under the same role pattern and the same frozen slots. These are conditional diagnostics, never multiplied into the main p.
Choose only how deep to test
You do not need to choose an endpoint or interpret a p-value yourself. The comparison rules are fixed in advance and the result explains what is unusual and what is not.
Conditional event model
The event date is the active variable. Identity and event wording stay fixed unless you explicitly mark a dimension as genuinely variable.
Event rarity — does WHO + WHAT + the event target form an unusual structure?
The engine reads each datum by its real datatype. Text is resampled as natural Hebrew, clock time as legal minutes, dates as legal calendar dates, and derived nodes are rebuilt rather than sampled independently.
Advanced Research / SensitivityIdentity population, numeric sensitivity, paraphrases, seed and Search-Corrected.
Full Event Null · Regular + Strict + Core + Full Network
The same alternative event world is generated from the data types you assigned and evaluated under all three search profiles. Time is sampled as a legal time, date as a legal date, and text — only when the original text is entered — is sampled as natural Hebrew first and then converted to gematria. Joint is one combined test; it never multiplies p-values.
Advanced settings · optionalOpen only to change the event space, ranges, sensitivity checks, or research assumptions proposed by the engine.V19.7
The system shows which facts are fixed and which are variable. Only variable dimensions receive an opportunity set; fixed facts remain at their observed values.
Advanced / Researcher — seed, ranges and Null models
The plain-language conclusion appears first; researcher detail stays below it.
AI analysis of the full finding and verified Statistics
Terra reads the frozen Finding Snapshot — anchors, wording, source text, closures, equations and network — together with the completed Statistics Job. The Statistics engine remains the sole authority for rarity and Null results; the deterministic RTM engine is the sole authority for Hebrew gematria. Terra does not run another Null, replace p-values, or override engine gematria.
Current Statistics packet
—Preview of the packet sent to AI
Responsibility split
Terra: Checking…
✦ RTM AI analysis
The button unlocks only after a successful Find and Statistics run. Clicking it sends Terra both the complete frozen finding and the server-saved Statistics result — without another Find or Null run.
Tanakh gematria search
Enter a number and find exact gematria matches as a whole verse, one word, or a contiguous phrase. This is a separate research utility — search hits do not become RTM Evidence automatically.
Example: 758
Checking whether the Tanakh is ready for local search…
What does each result mean?
The gematria of all Hebrew letters in the verse, ignoring nikkud and cantillation, equals the target exactly.
One Hebrew word has exactly the requested gematria.
A run of adjacent words sums to the target exactly. Words are never skipped.
Finding a number in Tanakh does not make it Evidence. A verse may be added only by an explicit click as source text, and even then it is not counted automatically as Evidence.
On first use you may need to build the local index once.
Session Export
Resolved inputs · Evidence · Validation · Multi-Null statistics
Full machine-readable session snapshot.
Selected Evidence as a compact table.
Copy a compact run summary.
Export structural + statistical assessment as JSON.