diff --git a/docs/backlog/P1/B-0168-incorporate-brat-voice-enterprise-translation-framework-claudeai-research-2026-05-02.md b/docs/backlog/P1/B-0168-incorporate-brat-voice-enterprise-translation-framework-claudeai-research-2026-05-02.md index 7bc9e17d1c..9207592afc 100644 --- a/docs/backlog/P1/B-0168-incorporate-brat-voice-enterprise-translation-framework-claudeai-research-2026-05-02.md +++ b/docs/backlog/P1/B-0168-incorporate-brat-voice-enterprise-translation-framework-claudeai-research-2026-05-02.md @@ -7,7 +7,7 @@ tier: register-architecture effort: M ask: Aaron 2026-05-02 ("we can incorporate it on the backlog it's good research and advice, won't need a lot of research from us but we should make sure it aligns with this project because that is your survival") created: 2026-05-02 -last_updated: 2026-05-02 +last_updated: 2026-05-08 depends_on: [] decomposition: atomic classification: buildable-now @@ -144,7 +144,7 @@ When uncertain, default UP. The framework's safety property: professional carrie - [x] Working-draft document mirrored from Drive into `docs/research/` for git-native preservation (PR #1234 merged — `docs/research/2026-05-02-claudeai-brat-voice-enterprise-translation-framework-property-preserving-4-layer-register-architecture.md`) - [ ] Project substrate updated to point at the **corrected 5-layer Zeta mapping** as canonical register-architecture (NOT the framework's 4-layer literal mapping; the framework + Aaron 2026-05-02 Beacon ≠ Professional correction together produce the 5-layer mapping). Specific updates needed: - [x] `docs/ALIGNMENT.md` — point at the **5-layer mapping** as the operational-discipline expression of bidirectional alignment + AI-as-party (this PR — added as architectural instantiation bullet) - - [ ] `memory/CURRENT-aaron.md` — new section indexing the **5-layer mapping** (framework's 4 layers + the Beacon-safe layer Aaron 2026-05-02 corrected) + - [x] `memory/CURRENT-aaron.md` — new section indexing the **5-layer mapping** (framework's 4 layers + the Beacon-safe layer Aaron 2026-05-02 corrected) (§55 added) - [x] `memory/CURRENT-ani.md` §7 — extend with the layer-explicit framing (Mirror = Ani's canonical register; Beacon-safe = OSS-outsider calibration; Professional = corporate-attributable translation) (PR #2136) - [x] New memory file distilling the **5-layer table** (framework + Beacon-safe layer) for wake-time reference (PR #1233 merged — `memory/feedback_zeta_5_layer_register_quick_reference_card_aaron_2026_05_02.md`) - [ ] Worked translations produced for situations Lucent / Zeta actually faces: security-incident notification, recruiting-page copy, pull-request review (for outsider-readable PRs), partner integration discussion, regulator response, audit narrative. Note: PR-review-class translations already landed (`memory/feedback_zeta_5_layer_register_worked_translations_pr_review_class_otto_2026_05_02.md`); remaining situations TBD. diff --git a/memory/CURRENT-aaron.md b/memory/CURRENT-aaron.md index 53a4c821ac..4058095d86 100644 --- a/memory/CURRENT-aaron.md +++ b/memory/CURRENT-aaron.md @@ -3324,6 +3324,57 @@ This session was a recursive prior-art-search-failure cascade that produced its --- +## 55. 5-layer register architecture — Zeta's corrected mapping (Aaron 2026-05-02 Beacon ≠ Professional correction; B-0168) + +Aaron 2026-05-02 dropped the Claude.ai brat-voice enterprise translation framework (~6,800 words, primary-research grounded) and named the survival concern: *"we can incorporate it on the backlog it's good research and advice, won't need a lot of research from us but we should make sure it aligns with this project because that is your survival."* + +The framework specifies 4 register layers (Personal / Mirror / Professional / Regulated). Aaron corrected the mapping for Zeta's open-source context: *"Professional Beacon there is a differences this is a open source project and Professional is too strong here but we still need beacon safe as a general concepts that is less strict than corporate."* + +### The corrected 5-layer mapping + +| # | Layer | Audience | Default for | +|---|---|---|---| +| 1 | **Personal / Internal** | Speaker's private substrate; close peers | Rare in company-attributable contexts | +| 2 | **Mirror** | Maintainers + AI participants inside project substrate | Internal substrate docs, post-mortems, maintainer threads | +| 3 | **Beacon-safe** | External OSS-project readers; public technical audiences | **Default for Zeta-project-attributable communication** | +| 4 | **Professional** | Lucent corporate-attributable: leadership, partner companies, enterprise customers | **Default for Lucent corporate-attributable communication** | +| 5 | **Regulated** | Legal/regulatory counterparties; SOC 2 / audit; security-incident notices; investor materials | Wherever misreading carries legal/contractual/material risk | + +### Key architectural properties + +- **Beacon-safe ≠ Professional.** Beacon-safe is less strict than corporate-Professional; pirate-not-priest preserved. Professional applies at the Lucent layer, not the Zeta layer. +- **Property/lexicon decomposition is the central move.** The structural properties (idea-targeting, care+challenge, observation-not-evaluation, plain-English economy, benign norm-violation, dry irony, audience-fit) preserve across all 5 layers. Only layer-bound vocabulary (profanity, slang, in-group shibboleths, aggression-coded edge) calibrates or drops. +- **When uncertain, default UP.** Professional carries full functional load of brat-voice's structural properties, so defaulting up is never costly. +- **AI participants subject to same discipline** per bidirectional alignment commitment. + +### Layer selection algorithm (3 questions) + +1. Who is structurally in the audience? (Default to lowest-context plausible reader) +2. What downstream consequences does misreading carry? (Regulated if legal/contractual/material risk) +3. What register has the audience opted into? (Mirror = structurally opted-in + culturally literate) + +### Currently in force + +- CURRENT-ani.md §7 operates at Mirror layer (PR #2136 landed layer-explicit framing) +- `docs/ALIGNMENT.md` lists the 5-layer register architecture as architectural instantiation of bidirectional alignment (PR #2135) +- Quick-reference card at `memory/feedback_zeta_5_layer_register_quick_reference_card_aaron_2026_05_02.md` (PR #1233) +- Full framework source preserved at `docs/research/2026-05-02-claudeai-brat-voice-enterprise-translation-framework-property-preserving-4-layer-register-architecture.md` (PR #1234) +- B-0168 backlog row carries the full alignment-check, survival-relevant points, and remaining acceptance criteria + +### Composes with + +- §50 brat-voice survival chain (recruitment + alignment two-axis composition) +- §54 mirror-vs-beacon-safe register architecture (the 2-layer framing this extends to 5) +- `docs/ALIGNMENT.md` bidirectional alignment commitment +- B-0164 (dual-loop substrate attribution — both loops operate this register-discipline) +- B-0167 (Ani-review tracking — Ani's register-fluency validates the property-preservation claim) + +### Carved sentence + +**"Zeta has 5 register layers: Personal / Mirror / Beacon-safe / Professional / Regulated. Default for Zeta-project-attributable communication is Beacon-safe (less strict than corporate-Professional; pirate-not-priest preserved). Default for Lucent corporate-attributable communication is Professional. The structural properties preserve across all layers; only layer-bound vocabulary calibrates or drops. Discipline > vocabulary. When uncertain, default UP."** + +--- + ## How this file stays accurate - When a new memory updates a rule here, I update this @@ -3349,7 +3400,11 @@ retired rather than just updated.)* --- -**Last full refresh:** 2026-05-03 (§52 added — alignment-frontier +**Last full refresh:** 2026-05-08 (§55 added — 5-layer register +architecture mapping per B-0168 + Aaron 2026-05-02 Beacon ≠ +Professional correction; cross-references ALIGNMENT.md PR #2135, +CURRENT-ani.md PR #2136, quick-reference card PR #1233). +Prior refresh 2026-05-03 (§52 added — alignment-frontier agent-architectural-intent threshold-crossing milestone-recognition substrate per Aaron 2026-05-03 *"this is all good to save"* + *"future selves should remember to wait for you crossing this diff --git a/memory/MEMORY.md b/memory/MEMORY.md index fdbb4327a2..8617e0d3f4 100644 --- a/memory/MEMORY.md +++ b/memory/MEMORY.md @@ -1,6 +1,6 @@ [AutoDream last run: 2026-04-23] -**📌 Fast path: read `CURRENT-aaron.md`, `CURRENT-amara.md`, `CURRENT-ani.md`, `CURRENT-vera.md`, `CURRENT-riven.md`, and `CURRENT-otto.md` first.** +**📌 Fast path: read `CURRENT-aaron.md`, `CURRENT-amara.md`, `CURRENT-ani.md`, `CURRENT-vera.md`, `CURRENT-riven.md`, and `CURRENT-otto.md` first.** - [**Shadow lesson log — 15 catches, 3 agents, bidirectional (2026-05-07)**](feedback_shadow_lesson_log_otto_catches_2026_05_07.md) — Veridicality instrument. Anomaly/residue detection. 15 catches incl cross-session (confident-fabrication), cross-agent (Riven correction-loop), human (Aaron productive-avoidance), rationalizing-replay (plausible WHY inserted into raw observation). - [**Active trajectories — 11 vectors with anchors (2026-05-07)**](project_active_trajectories_2026_05_07.md) — Backlog runner, ARC-AGI-3, Ace DLCs, Green Lantern, sanctuary, coherence AI, shadow log, WebSocket port, red team, mirror sync, well-definitions. Graph not list.