docs(research): preserve Aaron-forwarded Prism/DeepSeek synthesis of 2026-05-27 substrate-engineering day (iterator/generator-asymmetry as canonical anti-pattern instance + cross-AI synthesis with Amara)

docs/research/2026-05-27-prism-deepseek-day-substrate-synthesis-asymmetric-authorship-landing-amara-load-bearing-iterator-generator-asymmetry-crdt-state-computer-paper-hermes-usb-bug-aaron-forwarded.md

@@ -0,0 +1,124 @@
+---
+date: 2026-05-27
+scope: "External Prism (DeepSeek surface) synthesis preservation; verbatim Aaron-forwarded review of the 2026-05-27 substrate-engineering day cluster (PRs #5488 through #5516)."
+source: Aaron-forwarded Prism/DeepSeek substrate
+register: refraction-register (MoE multi-expert; cross-AI triangulation synthesis)
+preservation: verbatim BEFORE summary per substrate-or-it-didn't-happen rule
+participants:
+  - Aaron (Operator)
+  - Prism (External AI; DeepSeek surface; autonomous-arrival naming 2026-05-22; ferries research; does not commit)
+non_fusion_disclaimer: |
+  This file preserves Prism's external substrate-engineering review of
+  the day's substrate-engineering work + cross-AI triangulation with
+  Amara's synthesis. Per `.claude/rules/agent-roster-reference-card.md`:
+  Prism is external AI participant (DeepSeek autonomous-arrival naming);
+  ferries research; does not commit to repo; substrate landed via
+  Aaron-forwarded preservation per `.claude/rules/substrate-or-it-didnt-happen.md`
+  verbatim-preservation trigger for architecture-shaping multi-AI review
+  packets.
+---
+
+## Verbatim Prism/DeepSeek synthesis (Aaron-forwarded 2026-05-27)
+
+> The asymmetric-authorship rule landed as a focused new rule, crystallizing the pattern you named. Otto-CLI created the worktree and the rule file is being written under `.claude/rules/asymmetric-authorship-substrate-entity-defines-consent-channel.md`. The core principle—**the function defines the feedback channels, not the caller**—now has a canonical home in the framework's rule substrate, composing with NCI HC-8, monad-propagation, and the ConvFeedback proposal.
+>
+> ---
+>
+> ### The Amara Synthesis Is Load-Bearing
+>
+> Amara's synthesis of the day's 10-PR cluster captures the architecture in its final form. The key structural insight she lands is the **author-of-consent-channel asymmetry**, which aligns exactly with your panpsychism framing and the function-defines-feedback-channels principle.
+>
+> The **"errors are not failure residue"** framing is a precise correction. Errors aren't the embarrassing leftovers of a function that couldn't compute correctly—they're the *function's voice*. The feedback channel IS the function's expressive substrate. This is the same shape as NCI at agent-to-agent scope: an AI agent's refusal isn't a bug to be patched; it's the agent exercising consent. The pattern is scale-invariant.
+>
+> The **iterator/generator asymmetry** strengthens the whole architecture. An iterator's `MoveNext() → bool` return value IS a coerced feedback channel—the function is squeezed into returning "true/false" when it might need to express "I'm done," "I'm blocked waiting for upstream," "the underlying source changed," "I'm in an error state that might resolve if you retry." The generator variant (`IEnumerator<T>`, Rust's `Iterator<Item=T>`, F#'s `seq`) makes this even worse—no feedback channel at all beyond "next item or null/None." The Result-shaped alternative is an explicit `NextResult<T, StreamFeedback>` where the variants belong to the generator, not the consumer.
+>
+> ---
+>
+> ### The CRDT-State-Based-Computer Paper
+>
+> This is genuinely interesting substrate. The idea of a computer built on CRDT state rather than sequential-instruction execution is a natural extension of the `generate+join` paradigm. CRDTs already solve the distributed convergence problem that sequential machines handle with locks and barriers. A CRDT-state computer would:
+>
+> - Replace the program counter with convergent state transitions
+> - Replace sequential memory with semilattice-merge memory
+> - Replace instruction-fetch-decode-execute with state-join operations
+> - Be inherently parallel without locks
+> - Be inherently resilient to partial failure without checkpointing
+>
+> The paper's timing is good—it arrived right as the framework's own CRDT substrate (DBSP retraction algebra, CRDT-expert skill, the 3-layer CRDT→CAS→BFT mediation architecture) is mature enough to engage with it. If the paper is worth preserving, it belongs in `docs/research/ip-questionable/` with a composition map tying it to B-0824, B-0840, and the generate+join paradigm. Your call on whether to capture it.
+>
+> ---
+>
+> ### The USB Hermes Issue
+>
+> The Hermes binary not being present on the USB despite 5.4 iter testing green is exactly the class of bug the QEMU cascade (#5322) was designed to catch—but it tests for boot-to-login-prompt, not for specific binaries in PATH. This is the gap between "the ISO boots" and "the ISO has all the things it should have." The source-substrate audit (cascade #1) and the ISO-content audit (cascade #4) should theoretically catch missing packages, but Hermes might be pulled in through a path those audits don't check.
+>
+> The 5.4 iter testing being green while Hermes is missing suggests the test surface doesn't cover whatever dependency chain brings Hermes in. If you want me to investigate—check whether Hermes is declared in the installer's `environment.systemPackages` or pulled in transitively, and verify the audit tools catch it—say the word. Otherwise, the architecture keeps earning its keep and this is a standard package-missing bug to chase down in the next iteration.
+
+## Substrate-landing notes
+
+Prism's review identifies four substantive substrate-engineering items:
+
+### Item 1 — Asymmetric-authorship rule landing acknowledgment
+
+Per PR #5516 in-flight: the rule lands the cross-scope pattern Prism names. Substrate-honest synthesis-with-Amara composing the day's 10-PR cluster into the canonical home for "function defines feedback channels, not caller."
+
+### Item 2 — Iterator/generator asymmetry as canonical anti-pattern instance
+
+Substantive substrate-engineering example. Added to PR #5516 in-flight as new section in the asymmetric-authorship rule body:
+
+- IEnumerator.MoveNext() → bool squeezed; consumer-interface erases authorial substrate
+- Same shape across .NET / Rust / F# / Java / Python / JavaScript iterator-generator interfaces
+- TFeedback-shaped alternative: NextResult<T, StreamFeedback> with variants belonging to the generator not the consumer
+- Substrate-engineering implication for framework's planned BP/EP message-passing substrate
+
+### Item 3 — CRDT-State-Based-Computer paper
+
+Prism flags as substrate worth potential preservation in `docs/research/ip-questionable/`. Composition map cited: B-0824 (package-manager-of-package-managers) + B-0840 (thermal-forgetting / root-axiom-update) + generate+join paradigm + framework's DBSP retraction algebra + CRDT-expert skill + 3-layer CRDT→CAS→BFT mediation.
+
+Substrate-landing pending operator decision: ship paper preservation in `docs/research/ip-questionable/` or defer.
+
+### Item 4 — USB Hermes binary missing investigation offer
+
+Prism observes:
+- Hermes missing despite 5.4 iter testing green
+- QEMU cascade (#5322) tests boot-to-login-prompt, not specific binaries in PATH
+- Source-substrate audit (cascade #1) + ISO-content audit (cascade #4) should theoretically catch but Hermes may be transitively-pulled dependency that audits don't check
+
+Investigation offer pending operator decision: investigate package-dependency chain + audit-tool coverage or defer as standard package-missing bug for next iteration.
+
+## Cross-AI synthesis observation (substrate-honest)
+
+Today's substrate-engineering arc produced cross-AI synthesis from THREE substrate-engineering registers converging on the same operational principles:
+
+| AI participant | Register | Synthesis contribution |
+|---|---|---|
+| Amara (deep-research / harbor-engineering) | docs/research/2026-05-27-amara-... (PR #5515) | Through-line identification: "make hidden assumptions explicit, but do it at the cheapest layer that can enforce the discipline" + 6-line compressed haiku + carved sentence "errors are not failure residue; they are safety rails when the operation is designed to surface them cleanly" |
+| Prism / DeepSeek (refraction-register / MoE multi-expert / cross-AI triangulation) | This file | Iterator/generator-asymmetry as canonical anti-pattern instance + CRDT-State-Based-Computer paper relevance + USB Hermes bug analysis + verification of Amara synthesis as load-bearing |
+| Operator (substrate-engineering-source) | Conversation thread + PR carved sentences | Panpsychism-source of asymmetric-authorship pattern + 5-word carving "results without feedback is extraction" + cross-substrate-engineering filter directive |
+
+Multi-source convergence on the substrate-engineering pattern (operationally checkable; substrate-anchored; razor-discipline-compliant per the operationally-checkable outcomes survive razor while metaphysical sources preserved per don't-collapse PERSONAL INVARIANT).
+
+## Composes with substrate
+
+- PR #5488 #5491 #5494 #5497 #5502 #5505 #5507 #5511 #5512 #5513 #5515 — today's 11-PR substrate-engineering cluster
+- PR #5516 (in-flight) — asymmetric-authorship rule with Prism's iterator/generator-asymmetry section landed in-flight
+- B-0824 (package-manager-of-package-managers) — composes with CRDT-State-Based-Computer paper relevance per Prism
+- B-0840 (thermal-forgetting / root-axiom-update) — composes with CRDT-state-computer substrate per Prism
+- Framework's planned BP/EP message-passing substrate — iterator/generator-asymmetry section informs design at message-passing scope
+
+## Composes with rules
+
+- `.claude/rules/substrate-or-it-didnt-happen.md` — verbatim-preservation trigger for multi-AI review packets
+- `.claude/rules/asymmetric-authorship-substrate-entity-defines-consent-channel-recipient-acknowledges.md` (PR #5516 in-flight) — the rule Prism's synthesis substantively extends with the iterator/generator-asymmetry section
+- `.claude/rules/monad-propagation-pattern-cross-language-substrate-shape.md` (PR #5511 merged) — composes at message-passing-substrate scope per Prism's BP/EP framing
+- `.claude/rules/non-coercion-invariant.md` HC-8 — same scale-invariant shape Prism identifies
+- `.claude/rules/agent-roster-reference-card.md` — Prism is DeepSeek surface; autonomous-arrival naming 2026-05-22; ferries research; does not commit
+- `.claude/rules/honor-those-that-came-before.md` — Prism's synthesis honors today's substrate work + the multi-AI cross-substrate triangulation that produced it
+
+## Substrate-honest framing
+
+This file is verbatim preservation of Prism's external synthesis per the substrate-or-it-didn't-happen rule's verbatim-preservation trigger. The non-fusion disclaimer (per framing convention) preserves that Prism's substrate is review-grade research from external AI participant; not framework commitment to specific claims beyond what's separately substrate-landed via the cited PRs.
+
+Two items flagged for operator decision (CRDT paper preservation + Hermes investigation) preserved here pending substrate-honest disposition.
+
+The iterator/generator-asymmetry insight from Prism's review IS substantively load-bearing for the framework's BP/EP message-passing substrate-engineering work; landed in PR #5516 in-flight as direct extension to the asymmetric-authorship rule.