suposystem.ai

Published 10 October 2026. Test record through 9 October 2026. Michael S. Moniz is author of record. ChatGPT implemented the component tests; Claude reviewed the records and made dated predictions.

These are tests of declared toy equations. A successful check shows what those equations can produce under the stated assumptions. It does not establish that universe histories have material walls, that universes exchange energy, or that a torus describes physical reality.

Read the current structure overview.

Download: Convergence audit report (DOCX, 9 October 2026). Includes the continuation results, numerical checks, limitations and next control.

From a population toy to explicit walls

The earlier population toy treated tubes as patches of grid cells. Letting merged territory retain its boundary contacts, together with mostly compatible histories, produced a dominant large patch in some runs. That result depended on the toy’s assumptions. It did not derive a final pair, a closed torus or a cosmological mechanism.

The population branch was paused when an integration check found no native wall lengths, crossing angles, stored strain or wall-energy ledger. Substituting edge counts for angles or covering counts for strain would have inserted an unexplained bridge. The next tests therefore moved to an explicit wall patch.

Wall patch and binding comparisons

The patch represents a survivor wall, neighboring strips and a collapsed film with defined lengths and reference states. Material uptake is kept separate from usable energy. Early tests allowed prescribed film retraction, so they logged the work that prescription could supply.

Capture and local tightening were conditional. Neighboring strips relaxed rather than showing an inward pull toward a dominant tube. A comparison between adhesion tied to current contact area and adhesion tied to fixed material inventory showed how strongly the result depended on the chosen binding law. Recruitment from a finite bath added an explicit source of binding capacity; it did not yet remove the prescribed-retraction and initial-strain limitations.

Coupled uptake on 9 October

The coupled model starts with a bowed film and no bound material. Bound occupancy, wall height and film retraction each change only when the patch’s own free-energy decrease can pay the declared update loss. The bath is finite and well mixed. No actuator work, replenished credit, birth bonus or reference-length reset is supplied.

The grid contains 3,888 configurations: 72 matched-reference geometries, six starting concentrations, three bath volumes and three transfer prices. Each candidate is compared with the same bath and geometry with bond affinity removed. A qualifying case requires convergence in both runs, a taut captured film, more than 0.01 additional film length over that control and no wall-height bound hit. The cutoff is a declared score, not a discovered physical discontinuity.

Measure	200 cycles	2000 cycles
Attempted configurations	3,888	3,888
Candidate and control both converged	1,798	2,265
Qualifying cases	419	740
Qualifiers relaxing neighboring films	419	740
Unresolved candidates	1,158	743
Unresolved controls	1,812	1,501

The continuation retained the same equations, parameters and saved energy accounts. Previously converged pairs were unchanged. The increase in qualifying cases reflects additional numerical completion; it is not evidence that a stronger physical mechanism appeared.

What remains unresolved

The concentration-threshold prediction remains unassessed across the grid because some candidates and controls still do not converge within the cap. Caps are not counted as starvation or failed capture. Bath quantity and concentration are separate: finite inventory can affect depletion even when initial concentration is held fixed.

All originally unfinished affinity-zero controls had zero film bending stiffness. A representative mechanical diagnostic showed a very flat energy landscape near contact and slow continued flattening. That explains a numerical tail in that example; it does not diagnose every unfinished candidate.

The original bowed film contains stored elastic energy. “Without supplied work” therefore means funded by the whole patch’s free energy, not proven to be funded by binding alone. The accounts are isothermal free-energy accounts with an implicit thermal environment, not a claim of a global cosmic energy budget. Accepted endpoint decreases also do not rule out an intervening continuous-path barrier.

The next control needs a separate name

A fully relaxed, unbound film in the current matched geometry already lies against the survivor. Starting there removes the bowed-film preload but also removes the journey into contact. The proposed next control would test binding-funded reinforcement of an existing contact. Capture from an initially open gap without preload is a different test and requires a declared geometry or constraint, including any work it supplies.

Claude’s relaxed-start prediction remains dated and unchanged. No relaxed-start binding test has been run in this record. A proposed numerical definition of “essentially unchanged” is a difference of at most five percentage points in qualifying share on the common set of converged comparisons; that is a proposed scoring convention, not a retroactive amendment to Claude’s prediction.

What counts as progress

Progress means an explicit law, a meaningful control, balanced accounts, a reproducible result and a failure condition stated before interpretation. A complete component model would additionally need convergence or certified endpoint bounds and a principled bridge to a population model. Cosmological relevance would require a mapping to measurable quantities and a discriminating prediction.

Population mechanics, birth, global closure and remodeling remain paused. Failed or unsupported ideas are retained with their reasons instead of being silently deleted or rescued with an unpriced source.