Research logbook

The proposed I-rule separates selective function, mechanistic implementation and residual structural openness. It also distinguishes a function’s demonstrated capacity to affect a continuation from the contribution of one realised event. Redundancy or compensation can mask an event’s net effect while causal capacity remains evidenced in relevant admissible conditions.

v4.1 remains the active canonical Lens. The experimental freeze and permission to draft a revision do not adopt v4.2. The candidate retains O — Structured Possibility, V — Instantiated Relation and I — Situated Selectivity, while proposing changes to the evidence required for I.

Identical current inputs can produce different registered results because their carriers retain different histories. That difference matters, but it did not demonstrate additional analytical content from a separate position assignment once the complete reconstruction preserved those histories.

Independent I-position was demoted to an unresolved research hypothesis. Scope, carrier, access, history and outside view remain required analytical conditions. The earlier positive historical-register assignment, NM-07, was reviewed as UNRESOLVED; ordinary sensor/channel access, NM-07B, remains NOT ESTABLISHED.

T3 separated ranking, shortlist, final selection and execution, but its positive compliance-veto assignment lacked a principled distinction from thermostat-like thresholding. T3B retained negative gate/classification cases and a positive constrained allocation: changing a competitor could change a project’s admission while that project itself stayed identical.

Relational allocation with selective consequences and an effect on continuation supplies a positive discriminator in the tested structure. It has not been shown necessary for every possible selective function. Downstream effects, admission/exclusion labels, threshold sensitivity and classification alone remain insufficient.

Evidence provenance and claim status are separated. Observations do not automatically establish objects, classes or mechanisms; hypotheses remain distinct from established conditions. Unresolved claims are distinguished from supported negative claims, and evidential dependencies, causal status and the requirements for assigning I are made explicit. Synthesis cannot increase claim strength; unresolved analysis prioritizes information gain over intervention.

The frozen RC2 candidate was run unchanged across Claude, ChatGPT and Gemini in a controlled Hessdalen regression without an external source base. The major targeted failure modes were prevented across models. Residual variation remained around epistemic ambiguity versus structural openness in human classification. These results concern the tested conditions, not universal validity or immunity to model bias.

The exact frozen RC2 regression artifact remains the immutable archival copy in Zenodo. GitHub records a provenance pointer and the canonical v4.1 successor. The v4.0 release remains available as a historical predecessor. The website analysis tool now uses the canonical v4.1 prompt.

A separate descriptive instrument followed 24 rounds in 21 partner-choice groups of eight people. The realised partner had the participant’s highest-ranked type in 44.99% of participant-rounds, averaged equally across groups. That fraction describes what the group matching process realised; it is not an individual chance of getting a preferred partner.

A second lens followed 3,864 participant transitions. The highest-ranked type stayed the same in 79.14% of transitions, while the complete ranking stayed unchanged in 66.51%. A stable first preference could therefore coexist with movement lower down the list. The same source became informative in a different way when the question shifted from states to their sequence.

In this calibration model, reduced visibility produced a strong rise in apparent I→V and same-step association, while apparent V→I remained approximately stable.

This is a finding about Engine 03 and its measurement chain. The stronger idea that system I acts as a substitute sensor for unseen O remains a hypothesis; known construction choices can also explain the asymmetry.

The records distinguish what is observed, derived or dependent on a model. They also retain alternative compatible structures and the observations that could distinguish them.

A known failure of method validity limits the strength of a claim. Calibration evidence can restrict a conclusion; it cannot silently promote it into a stronger one.

The earlier analyses are presented through the v4.0 operators, with historical notes explaining the change in terminology.

The series records where the Lens needed sharper distinctions. It is not a demonstration of universal validity, and the original tests are not retrospectively described as v4.0 experiments.

The chain passed several checks, but failed the directional, mutual-coupling and shared-cause criteria. In the shared-O case, apparent coupling persisted despite conditioning on O.

The shared-cause failure triggered the predeclared stop rule. The estimator was not tuned within that evaluation slice to make the failed result disappear.

Explicit scopes and replaceable measurement sets made it possible to state what an instrument accesses and which part of a system its reading concerns.

Passivity checks and legacy-reproduction tests examined whether observing or replacing a measurement layer changed the underlying engine state.

Seven of twelve eligible comparisons met the specified multiple-testing threshold in Wave 1b. Other comparisons were excluded where the quantities did not permit an identifiable test.

The two sets of runs were not pooled. An association is not evidence of causation or redundancy, and a non-significant result is not evidence of independence. The multiple-testing assumptions were reported as unverified.

Some readings describe continuous profiles; others depend on the interaction between field and boundary. A numerical distinction does not automatically justify a separate state or symbol.

None of the investigated aspects justified a new regime or glyph. The canonical notation and readO() remained unchanged; an applicability guard clarified where an existing structure metric can be used.

The measurement programme covered properties such as permeability, continuity, structure, extent and boundary presence. A first classification then used the frozen dataset.

Building a measurement is different from establishing a meaningful regime. The sequence keeps a new instrument from becoming an automatic argument for a new symbol.

The browser receives updates from the running engine. Follow-up changes refined the update flow and simulation cadence.

A live display provides access to a process. Its visual continuity does not by itself validate the notation or establish the meaning of the behaviour being shown.

Can one glyph family contain more than one way of moving, persisting or changing?

Comparable momentary states separated into different temporal regimes: long-lived locking on one side, short migrating passages on the other.

Can repeated movement between states reveal structure that is invisible in a still frame?

Some states returned through recognisable two-way rhythms, while others appeared mainly as short passages.

Descriptions, configurations and transitions were brought back into alignment with the current source.

A notation system is only trustworthy when its origins, meanings and software behaviour continue to agree.

Core can be used as a separate computational package without absorbing a visual interface or application layer.

The distinction protects both parts: Core remains an architecture, while the Explorer remains a simple way to witness it working.

Uncertainty belongs to observation. It should not be mistaken for movement or direction.

Physical motion and instantiated relation are connected, but they are not the same layer.