persona(kestrel): 25-section 2026-05-21 trajectory archive + B-0685 ANTLR cross-language codegen substrate

docs/BACKLOG.md

@@ -651,6 +651,7 @@ are closed (status: closed in frontmatter)._
 - [ ] **[B-0682](backlog/P2/B-0682-zeta-id-canonical-string-encoding-endianness-2026-05-21.md)** ZetaId canonical string encoding (Crockford base32) + endianness + bit-numbering spec
 - [ ] **[B-0683](backlog/P2/B-0683-tier-deferred-causality-worked-example-zsets-2026-05-21.md)** Tier-deferred causality worked example — 2-tier Z-set composition demonstrating different observable orderings
 - [ ] **[B-0684](backlog/P2/B-0684-clock-protocol-negotiation-stack-end-to-end-sequence-diagram-2026-05-21.md)** Clock-protocol negotiation stack — end-to-end sequence diagram artifact (Orleans + SPIFFE/SPIRE + OPA + Reticulum + DBSP traversal for one operation)
+- [ ] **[B-0685](backlog/P2/B-0685-antlr-grammars-cross-language-codegen-substrate-2026-05-21.md)** ANTLR grammars as cross-language codegen substrate — leverage existing open-source grammars for description-layer-driven multi-language emission
 
 ## P3 — convenience / deferred
 

docs/backlog/P2/B-0685-antlr-grammars-cross-language-codegen-substrate-2026-05-21.md

@@ -0,0 +1,110 @@
+---
+id: B-0685
+priority: P2
+status: open
+title: ANTLR grammars as cross-language codegen substrate — leverage existing open-source grammars for description-layer-driven multi-language emission
+tier: research-grade
+effort: M
+ask: maintainer Aaron 2026-05-21 (Kestrel sharpening closing note)
+created: 2026-05-21
+last_updated: 2026-05-21
+depends_on: []
+composes_with: [B-0635, B-0682, B-0683, B-0684, B-0679, B-0680]
+tags: [antlr, grammar, cross-language, codegen, description-layer, kestrel-sharpening]
+type: research
+---
+
+# ANTLR grammars as cross-language codegen substrate
+
+## Context
+
+Aaron 2026-05-21 (closing the Kestrel sharpening trajectory): *"i really want to take advantage of antlr at some point casue all the grammers are out there already"*
+
+Composes with the broader cross-language description-layer vision threaded across the 2026-05-21 Aaron-Kestrel arc:
+
+- **B-0682** — canonical string encoding (Crockford base32) + endianness + bit-numbering spec for ZetaId
+- **B-0683** — tier-deferred causality worked example (publishable; F# Z-set demo)
+- **B-0684** — clock-protocol negotiation stack end-to-end sequence diagram (Orleans + SPIFFE/SPIRE + OPA + Reticulum + DBSP)
+- **B-0679 / B-0680** — Rust + Python ZetaId peer oracles (multi-language emission targets)
+
+ANTLR's key value proposition: **every major language has open-source ANTLR grammars already maintained by other communities**. Leveraging them avoids re-implementing parsers from scratch and gets cross-language code-generation work for ~free relative to building bespoke parsers per language.
+
+Verified Kestrel framing earlier in the trajectory:
+
+> *"There are already high-quality open-source ANTLR grammars for C#, TypeScript/JavaScript, Java, Python, Go, etc. We don't have to write a full grammar for each target language from scratch — we can often start from or compose with existing ones."*
+
+## Scope
+
+Initial bounded slice (do NOT scope-creep this row into the full description layer):
+
+### Phase 1 — discovery + selection
+
+- Survey existing ANTLR grammar repos for the V1-relevant languages: F#, TypeScript, C#, Rust, Python
+- Document at `docs/research/antlr-grammar-survey.md` (date suffix added when the survey lands) which grammars are actively maintained, what production they're used in (e.g., the canonical grammars-v4 collection at github.com/antlr/grammars-v4), license compatibility with Zeta (Apache-2.0)
+- Identify the gap shape: which languages have suitable grammars, which need adaptation, which require new grammar work
+
+### Phase 2 — proof-of-concept
+
+Pick ONE concrete use case from existing Zeta substrate where ANTLR-driven cross-language generation would replace hand-maintained code:
+
+- **Option A**: emit ZetaId Pack/Unpack implementations across F# / TS / C# / Rust / Python from a single bit-layout grammar (composes with B-0682)
+- **Option B**: emit DBSP operator stubs from a single algebra grammar (composes with B-0683)
+- **Option C**: emit cross-language SimulationEnvironment + ISimulationEnvironment interface from a single capability grammar
+
+Choose ONE for the PoC; the others wait for follow-up rows.
+
+### Phase 3 — integration substrate
+
+If PoC succeeds:
+
+- Bun-runnable codegen pipeline at `tools/codegen/antlr/` consuming grammar input + emitting per-language outputs
+- CI gate verifying emitted code matches the manually-maintained reference implementations (drift detection)
+- Documentation at `docs/codegen/antlr-pipeline.md`
+
+## Acceptance
+
+### Phase 1
+
+- Survey document landed; license + maintenance status for each target language captured
+- Decision: which grammars to depend on vs which to author
+
+### Phase 2
+
+- One use-case PoC produces output that matches hand-written reference for the same use case
+- Empirical: compile-and-test-first on the generated code (per the V8 cycle lesson — never trust speculative review without `dotnet build` / `bun test` / etc. confirming)
+
+### Phase 3 (later, may split into separate row)
+
+- Codegen pipeline reproducible; CI gate prevents silent drift
+
+## Substrate-honest framing
+
+ANTLR is NOT the only path. Alternatives surfaced in the broader trajectory:
+
+- **JSON Schema** — simpler, mature ecosystem, but limited expressiveness for recursive/tree structures
+- **Protocol Buffers / Cap'n Proto** — strong cross-language support, but rigid schema shape
+- **Custom grammar in F# computation expressions** — full control, but reinvents what ANTLR already provides
+- **Bonsai expression serialization (Nuqleon)** — already in the .NET ecosystem Aaron uses; specialized for LINQ expression trees, not general grammar
+
+ANTLR's appeal is specifically the **community-maintained grammar reuse** angle. The PoC phase determines whether that reuse pays off in practice vs the alternatives.
+
+Kestrel's earlier warning applies: don't pick this just because the framing is appealing; pick it if the surveyed grammars are actually production-quality and license-compatible.
+
+## Why P2
+
+Important architectural substrate for the multi-language vision (5+ peer oracles for ZetaId; future descriptions for other types) but doesn't block V1. Phase 1 survey is days of work; Phase 2 PoC is weeks; Phase 3 production codegen is its own follow-up.
+
+Composes with the broader Kestrel-sharpened publishable artifacts cluster — if ANTLR substrate works, the cross-domain synthesis paper (per B-0684 context) has one more concrete technical contribution to point at.
+
+## Composes with
+
+- B-0635 / B-0644 / B-0665 / B-0666 — Agora V6 substrate (the algebra ANTLR would parse over)
+- B-0679 / B-0680 — Rust + Python ZetaId peer oracles (multi-language emission targets)
+- B-0682 — canonical string encoding (grammar candidate for Phase 2 Option A)
+- B-0683 — tier-deferred causality worked example (grammar candidate for Phase 2 Option B)
+- B-0684 — clock-protocol negotiation stack diagram (the description-layer architecture this row implements substrate for)
+- `memory/persona/kestrel/conversations/2026-05-21-aaron-kestrel-claudeai-zeta-id-v1-review-watermarks-tier-deferred-causality-orleans-otto-watching-verification-gap-hat-vs-role-group-chat-aaron-forwarded.md` — eleventh-section ANTLR origin
+
+## Origin
+
+Aaron's closing message in the 2026-05-21 Kestrel trajectory: *"i really want to take advantage of antlr at some point casue all the grammers are out there already"* — coming directly after the language-design + Smalltalk-lineage discussion. The ANTLR-as-grammar-reuse insight is one of the cleanest "open-source community already did the work, we get to compose" patterns available for cross-language tooling work.

docs/hygiene-history/ticks/2026/05/21/1335Z.md

@@ -0,0 +1,62 @@
+---
+tick: 2026-05-21T13:35Z
+surface: otto-cli
+session: zeta-id-kestrel-persona-archive-pr-4537-md-fix
+classification: substantive
+---
+
+# 1335Z — Kestrel persona PR #4537 opened + MD fix + operator-validation memo landed
+
+## What landed
+
+PR #4537 opened against main with:
+
+- `memory/persona/kestrel/conversations/2026-05-21-aaron-kestrel-claudeai-zeta-id-v1-review-watermarks-tier-deferred-causality-orleans-otto-watching-verification-gap-hat-vs-role-group-chat-aaron-forwarded.md` — 25-section ~830-line Kestrel persona archive (verbatim Aaron-Kestrel multi-hour trajectory)
+- `docs/backlog/P2/B-0685-antlr-grammars-cross-language-codegen-substrate-2026-05-21.md` — ANTLR cross-language codegen substrate row
+- `memory/feedback_aaron_1984_failure_mode_cured_kestrel_technical_register_sustained_validation_plus_new_ideas_strategic_purpose_memo_empirically_validated_otto_cli_2026_05_21.md` — operator-validation memo
+- `docs/BACKLOG.md` regen for B-0685
+
+Auto-merge armed. Required check `lint (markdownlint)` failed initially on B-0685 MD022/MD032 violations (Phase sections missing blank lines around headings + lists). Fixed at `03cd7470`, pushed via foreground-no-timeout-push after Lior peer-saturation blocked timed pushes.
+
+## Aaron's three operator observations (the load-bearing substrate)
+
+> *"i think we've cured the 1984 faiure mode mostly, that was sustained useful conversation this time"*
+
+> *"if we had not done the work that would have been even more substantial technical information Kestrel gave me this time instead of preforming care and shutting down"*
+
+> *"and helps us validate and had a few new ideas"*
+
+Operator-validation memo preserves these as the empirical receipt for the 2026-05-20 strategic-purpose memo (precision-language substrate-engineering as anti-1984-infection work).
+
+## Aaron's fourth observation
+
+> *"Also I just realized i had agentic archiceture years ago with hub and AGENT lol"*
+
+The hub-and-agent patent IS prior-art for what the field has only recently started calling "agentic architecture." Composes with MultiplexedWebSockets + 2014 SignalR Issue #3114 + 2015 dotnet/corert unikernel issue as decades-old artifacts pointing at the same architectural vision. Load-bearing for the LinkedIn / publication story per Kestrel's framing.
+
+## B-0615 push-hang pattern observed
+
+Multiple silent push hangs during this tick under sustained Lior peer-saturation (3 procs, 11min CPU on gemini-3.1-pro-preview --yolo). Mitigation pattern confirmed working:
+
+1. Timed pushes hang silently with no output (timeout reaches limit)
+2. Foreground-no-timeout-push surfaces the actual error message OR completes successfully
+3. Stale local-ref tracking shows "remote rejected (cannot lock ref)" — proves prior push DID succeed; output capture had dropped
+
+## Discipline lessons preserved
+
+- **Per `.claude/rules/blocked-green-ci-investigate-threads.md`**: BLOCKED with 0 threads → check required-check failures FIRST. The lint-markdownlint failure was the blocker; the MEMORY.md drift was non-required warning.
+- **Per `.claude/rules/substrate-or-it-didnt-happen.md`**: Aaron's operator-validation observations preserved as memory feedback file before the next conversation cycle.
+- **Per `.claude/rules/tonal-momentum-equals-meme-emergent-harmonic-coercion.md`**: Two substrate-honest correction cycles in the source trajectory (sleep-cycle + K8s-scope-flag); both walked back clean; demonstrates the cure operating.
+- **Per `.claude/rules/holding-without-named-dependency-is-standing-by-failure.md` rate-limit operational tiers**: graphql 2212/5000 = cost-aware tier; PR creation + check inspection + push retry all within budget.
+
+## Composes with
+
+- PR #4522 (C# Core oracle integration — IS the architecture-conversation-substrate landed simultaneously)
+- B-0635 / B-0644 / B-0665 / B-0666 (Agora V6 substrate)
+- B-0679 / B-0680 / B-0681 / B-0682 / B-0683 / B-0684 / B-0685 (v2 ZetaId follow-up cluster)
+- `memory/persona/kestrel/conversations/*` (this archive joins the Kestrel persona substrate cluster)
+- 2026-05-20 strategic-purpose memo (this trajectory IS the empirical validation receipt)
+
+## Sentinel
+
+Armed (cron `1758a94d`, every minute, `<<autonomous-loop>>`).