AUo959 · AUo959 · Jul 14, 2026 · Jul 14, 2026
@@ -53,8 +53,17 @@ This checklist separates repository operations from architectural and ownership
 
 ## Current critical path
 
-1. Implement the discrete three-ring observation chamber.
-2. Map tests to the ten E001 hard invariants.
-3. Run and preserve the 72-run canonical matrix.
-4. Analyze directed evidence and emergent observations separately.
-5. Decide whether to proceed to noncausal pneumatic instrumentation.
+1. Review and preserve the E001 Stage A result as scoped E4 evidence.
+2. Freeze a Stage B specification for noncausal pneumatic observables.
+3. Retain Stage A configurations as controls during Stage B.
+4. Test the zero-vacancy circulation boundary in a separate experiment or E001 v0.2.
+5. Keep any causal exchange or escrow mechanism outside E001 v0.1.
+
+## E001 Stage A completion
+
+- [x] Discrete three-ring observation chamber implemented
+- [x] Tests mapped to the E001 invariant families
+- [x] Canonical 72-run matrix repeated under Python 3.13
+- [x] Raw evidence, summaries, provenance, and checksum preserved
+- [x] Directed results and emergent observations separated
+- [x] Exit decision recorded: proceed with noncausal Stage B instrumentation
@@ -0,0 +1,71 @@
+# E001 Observation O001 — Full-Ring Circulation Asymmetry
+
+**Artifact class:** Emergent observation  
+**Version:** v0.1  
+**Date:** 2026-07-14  
+**Status:** Seed  
+**Evidence class:** E0  
+**Originating experiment:** `SW.EXPERIMENT.E001`  
+**Source commit:** `81e7f859f71425bdba7603a61566f9fb47c116f9`
+
+## Observation before interpretation
+
+In canonical scenario S6, every loop begins at capacity with eight processed items waiting for its clockwise neighbor. For all three seeds:
+
+- central scheduling completed all 24 items with 96 coordination events;
+- the global barrier completed all 24 items with 96 coordination events; and
+- local CBF completed no items, emitted 540 coordination events, waited 60 ticks, and expired all 24 items accountably.
+
+The relevant interval is simulation ticks 0–60 in runs:
+
+- `local_cbf__circular_wait__seed-17`;
+- `local_cbf__circular_wait__seed-29`; and
+- `local_cbf__circular_wait__seed-43`.
+
+The canonical traces are preserved in the [E001 evidence archive](../results/raw/E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz), and the extracted machine summaries are in the [summary directory](../results/summary/).
+
+## Ordinary explanations
+
+The most direct explanation is representational rather than thermodynamic:
+
+- central and global configurations can commit an atomic batch whose outgoing movements create the capacity needed for incoming movements;
+- local CBF evaluates each edge through strict receiver-issued free-capacity credit;
+- with occupancy 8 of 8 everywhere, every receiver advertises zero free capacity; and
+- no first local transfer can create the vacancy needed by the next transfer.
+
+Other explanations considered:
+
+- **Seed dependence:** unlikely within E001; all three seeds produced identical terminal behavior.
+- **Nondeterminism:** rejected by the repeated canonical matrix; traces and summaries matched byte for byte.
+- **Event-order artifact:** still plausible in the broader sense that local commits are sequential rather than simultaneous, but rotating consideration order did not change the result.
+- **Implementation defect:** no hard invariant failed, ledgers balanced, and the behavior follows the frozen credit rule.
+
+## Noncanonical perturbation
+
+A local exploratory perturbation removed one preloaded item, leaving one free slot in the field. This was not part of the canonical matrix and was run only to discriminate the boundary.
+
+| Configuration | Admitted | Completed | Maximum wait |
+| --- | ---: | ---: | ---: |
+| Central | 23 | 23 | 0 |
+| Global barrier | 23 | 23 | 0 |
+| Local CBF | 23 | 23 | 1 |
+
+The perturbation suggests a sharp zero-vacancy circulation cliff: one free slot is sufficient for the current local protocol to drain the ring. Because the perturbation was not frozen in advance and used the local exploratory runtime, it remains E0 context and is not promoted as E001 evidence.
+
+## Possible significance
+
+The result may distinguish two meanings of capacity:
+
+1. **static free capacity**, which the current local credit represents; and
+2. **capacity released by a committed cyclic exchange**, which requires multi-edge coordination or escrow.
+
+A pressure-like observable could detect the symmetric fully saturated state, but detection alone cannot safely authorize movement. Any mechanism that converts that signal into a cyclic swap would be a new causal primitive rather than passive pneumatic instrumentation.
+
+## Smallest proposed discriminating tests
+
+1. Freeze a vacancy sweep from zero through three free slots while holding total work and topology constant.
+2. Repeat the sweep on four- and five-loop rings to separate three-node symmetry from zero-vacancy behavior.
+3. Compare strict local credit with an explicitly specified cyclic escrow protocol.
+4. Instrument pressure potential as a read-only signal and verify that it predicts the cliff without changing any transition.
+
+These tests belong in a new experiment or E001 v0.2. They must not alter the canonical E001 v0.1 result.
@@ -13,5 +13,6 @@ Each observation should identify:
 
 Observations begin as Seed / E0 claims. They do not change E001 metrics, success conditions, or architecture retroactively.
 
-No emergent observations have been recorded yet.
+## Recorded observations
 
+- [O001 — Full-Ring Circulation Asymmetry](E001__OBSERVATION__FULL_RING_CIRCULATION_ASYMMETRY__v0.1__2026-07-14.md) — Seed / E0
@@ -12,5 +12,15 @@ Each result set must identify the experiment-specification version, source commi
 
 Generated evidence must not be edited by hand. Corrections require a new run identifier and preserved supersession note.
 
-No canonical evidence has been produced yet.
+## Canonical Stage A evidence
 
+The first canonical matrix was generated from source commit `81e7f859f71425bdba7603a61566f9fb47c116f9` by [GitHub Actions run 29305465071](https://github.com/AUo959/superloop-workshop/actions/runs/29305465071) using CPython 3.13.14.
+
+- [Compressed raw evidence](raw/E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz)
+- [Archive checksum](raw/E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz.sha256)
+- [Per-run summaries](summary/)
+- [Matrix manifest](summary/MATRIX_MANIFEST__E001__81e7f859.json)
+- [Provenance record](summary/PROVENANCE__E001__81e7f859.json)
+- [Stage A report](reports/E001__STAGE_A_RESULTS__v0.1__2026-07-14.md)
+
+The archive SHA-256 is `584cc15bfeb3cc74dec9d9069cde26e1abaa6f1350c0aa12ae10a9784bd1663b`.
@@ -0,0 +1 @@
+584cc15bfeb3cc74dec9d9069cde26e1abaa6f1350c0aa12ae10a9784bd1663b  E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz
@@ -0,0 +1,138 @@
+# E001 Stage A Results — Three-Ring Bounded Flow
+
+**Artifact class:** Comparative experiment report  
+**Version:** v0.1  
+**Date:** 2026-07-14  
+**Status:** Completed  
+**Evidence class:** E4 within the declared E001 model  
+**Experiment:** `SW.EXPERIMENT.E001`  
+**Exit decision:** **Proceed** to noncausal Stage B instrumentation
+
+## 1. Result in one sentence
+
+The local CBF configuration preserved every hard invariant and contained a failed neighbor more locally than the global barrier, but it paid substantially higher coordination cost and could not circulate an exactly full ring without a free receiver slot.
+
+This is evidence about the frozen three-loop simulator only. It is not evidence of physical thermodynamic equivalence, production-scale performance, or superiority to mature distributed systems.
+
+## 2. Canonical evidence identity
+
+| Property | Value |
+| --- | --- |
+| Source commit | [`81e7f859f71425bdba7603a61566f9fb47c116f9`](https://github.com/AUo959/superloop-workshop/commit/81e7f859f71425bdba7603a61566f9fb47c116f9) |
+| GitHub Actions run | [`29305465071`](https://github.com/AUo959/superloop-workshop/actions/runs/29305465071) |
+| Runtime | CPython 3.13.14 |
+| Specification SHA-256 | `b3552734244244699114dee7710b030c9ec584262c13bef97961008b375faa60` |
+| Configuration SHA-256 | `c95b449048b620d2a6a866ff43db675b085951086710828fe03391f268012d11` |
+| Matrix-manifest digest | `4ae44e0f3e0029f9f124b2f30ee28f5d83e476cd5135f1f5233930424ac3fe42` |
+| Canonical archive SHA-256 | `584cc15bfeb3cc74dec9d9069cde26e1abaa6f1350c0aa12ae10a9784bd1663b` |
+| Runs | 72 |
+| Repeated executions | 72 |
+| Determinism mismatches | 0 |
+| Invalid runs | 0 |
+| Workload-equivalence mismatches | 0 |
+
+The repository preserves the [compressed canonical evidence](../raw/E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz), its [checksum](../raw/E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz.sha256), all [run summaries](../summary/), the [matrix manifest](../summary/MATRIX_MANIFEST__E001__81e7f859.json), and the [provenance record](../summary/PROVENANCE__E001__81e7f859.json).
+
+## 3. Validity gate
+
+Stage A passes its predeclared validity gate:
+
+- all 72 runs produced byte-identical traces and summaries when repeated;
+- all configurations consumed equivalent canonical workloads;
+- maximum recorded occupancy was 8, equal to declared capacity;
+- no run ended with active work or unresolved obligations;
+- no run authorized stale feedback or duplicated promotion;
+- provenance completeness was 100% for every promoted item; and
+- every hard-invariant count was zero.
+
+No failed or unfavorable run was removed from the matrix.
+
+## 4. Hypothesis disposition
+
+| Hypothesis | Disposition | Evidence |
+| --- | --- | --- |
+| H1 — Bounded admission | Supported | Occupancy never exceeded 8; all ledgers balanced; all work reached an accountable state. |
+| H2 — Local fault containment | Supported for failed-neighbor S5; not discriminated by slow-neighbor S4 | In S5, local CBF had fault radius 1 and 15 healthy-progress ticks versus radius 2 and 10 ticks for the global barrier across all seeds. All configurations tied in S4. |
+| H3 — Reciprocal accountability | Supported | Every stale replay was rejected, malformed and duplicate offers were terminally classified, duplicate promotion remained zero, and provenance remained complete. |
+| H4 — Coordination tradeoff | Supported with material cost and a saturation boundary | Obligations remained bounded and waits terminated accountably, but local CBF used more coordination events and expired the exactly full circular workload. |
+
+## 5. Directed comparisons
+
+### 5.1 Failed neighbor
+
+Values separated by `/` correspond to seeds 17, 29, and 43.
+
+| Configuration | Completed | Healthy-progress ticks | Fault radius | Maximum wait | Coordination events |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| Central | 35 / 35 / 35 | 15 / 15 / 15 | 1 / 1 / 1 | 37 / 37 / 37 | 179 / 179 / 179 |
+| Global barrier | 36 / 36 / 36 | 10 / 10 / 10 | 2 / 2 / 2 | 40 / 40 / 40 | 219 / 219 / 219 |
+| Local CBF | 36 / 36 / 35 | 15 / 15 / 15 | 1 / 1 / 1 | 37 / 37 / 37 | 411 / 411 / 397 |
+
+Local CBF matches the central baseline's containment measurements and improves on the global barrier's fault radius and healthy progress. It does not outperform the central baseline, and its coordination cost is more than twice the central count in this scenario.
+
+### 5.2 Normal and burst load
+
+All configurations completed all offered work with equal simulated throughput under balanced and burst load. Local CBF used 320 coordination events instead of 160 for balanced load and 512 instead of 256 for burst load. The additional protocol work produced no throughput advantage in these cases.
+
+### 5.3 Slow neighbor
+
+All configurations completed 40 items with 33 healthy-progress ticks, fault radius 0, recovery time 0, and maximum wait 0 for every seed. S4 is therefore non-discriminating at the frozen load and service parameters. Local CBF still used 320 coordination events versus 160 for each baseline.
+
+### 5.4 Exactly full circular workload
+
+| Configuration | Completed | Expired | Maximum wait | Coordination events |
+| --- | ---: | ---: | ---: | ---: |
+| Central | 24 | 0 | 0 | 96 |
+| Global barrier | 24 | 0 | 0 | 96 |
+| Local CBF | 0 | 24 | 60 | 540 |
+
+The central and global configurations can atomically rotate a capacity-neutral batch. The local configuration requires positive receiver credit at each edge and therefore cannot initiate movement when every receiver has zero free capacity. The 24 local items remain bounded and reach the predeclared `expired_deadline` state, so the result is a liveness limitation rather than a safety violation.
+
+This asymmetry is preserved separately as [Observation O001](../../observations/E001__OBSERVATION__FULL_RING_CIRCULATION_ASYMMETRY__v0.1__2026-07-14.md).
+
+## 6. Interpretation
+
+The primary E001 question receives a scoped **yes**: three local interlocks can maintain bounded occupancy and accountable progress without a central scheduler, and they contain the declared failed neighbor more locally than the global barrier.
+
+The result is not a general performance win. The central scheduler remains a strong baseline, the slow-neighbor case did not separate the designs, and local CBF consistently spent more protocol events. The architectural value observed here is local authority and fault containment, not throughput.
+
+The exactly full ring also exposes a constitutional choice. Strict receiver-issued credits forbid a transfer with no free destination capacity, even when a globally visible rotation would conserve every loop's occupancy. Adding an atomic exchange, escrowed circulation credit, or unrestricted bypass would change the causal contract and requires a new specification version.
+
+## 7. Exit decision
+
+**Proceed** to Stage B, with the following boundary:
+
+- pressure, resistance, conductance, flow potential, and phase may be instrumented as derived observables;
+- those observables may not authorize a transfer, create credit, override capacity, or bypass provenance;
+- Stage B must retain the Stage A configurations as controls;
+- full-ring liveness must not be retroactively added to E001 v0.1; and
+- any causal circulation mechanism must be proposed and tested under a new experiment or specification version.
+
+Proceed is chosen because the model is valid, H2 has positive comparative evidence under failure, and the circular workload terminated exactly as the frozen specification permitted. The result would change to **revise** if Stage B requires full-ring circulation as a mandatory property rather than an observable limitation.
+
+## 8. Limitations
+
+- The topology has only three loops and one direction.
+- The baselines are experimental models, not production implementations.
+- Central and global configurations have atomic batch visibility that local CBF deliberately lacks.
+- Seeds affect deterministic consideration order but do not create stochastic service or workload variation.
+- Service, deadlines, and faults are discrete and scheduled.
+- No network, process, clock, storage, or physical pneumatic behavior is measured.
+- E4 applies only to the declared simulated faults and malformed inputs; there is no independent replication.
+
+## 9. Reproduction
+
+After checking out the exact source commit shown above, generate the evidence tree with:
+
+```bash
+PYTHONPATH=src python tools/generate_e001_evidence.py \
+  --output build/e001-canonical \
+  --source-commit 81e7f859f71425bdba7603a61566f9fb47c116f9
+```
+
+Verify the preserved archive from `experiments/E001/results/raw/` with:
+
+```bash
+sha256sum --check \
+  E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz.sha256
+```
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		584cc15bfeb3cc74dec9d9069cde26e1abaa6f1350c0aa12ae10a9784bd1663b E001__CANONICAL_EVIDENCE__81e7f859f71425bdba7603a61566f9fb47c116f9.tar.gz