Session Decisions

SD-328 Technical debt is an ever-present threat — systematic codebase review as interview preparation ACTIVE → SD-327 The Pit development paused for 6 weeks — interview preparation begins 14 March 2026 SUPERSEDED by SD-328 → SD-326 Multi-agent orchestration (gastown) not adopted for this workflow STANDING → SD-325 git stash is forbidden, no exceptions PERMANENT → SD-324 Inter-container communication via Docker named volumes COMPLETE → SD-323 Switch midget container base from ubuntu:24.04 to debian:bookworm-slim COMPLETE → SD-322 [midget-castle] **First build trajectory for midgets.** Three phases: (A) get an agent operating inside the container - … **ACTIVE** → SD-321 [signal-killed] **Operator verbatim:** "Signal has no signal. Kill it." Signal notation abandoned as a governance … **PERMANENT** → SD-320 [signal-adversarial-test] **Operator verbatim:** "I have serious concerns about Signal compression, and that the same or … **IN PROGRESS** → SD-319 [no-em-dash-no-emoji] **No em-dashes, for any reason, ever. No emojis, for any reason, ever.** Applies to all agent … **STANDING ORDER** → SD-318 [darkcat-alley] **Darkcat Alley — standardised 3-model cross-triangulation process.** Named, lexified (v0.25), … **STANDING (this run)** → SD-317 [qa-sequencing-data-products] **QA sequencing produces three data products: (1) Cross-Model Triangulation — 3 … **STANDING (this run)** → SD-316 [backref-density] **Backref density policy: 9 mechanisms adopted to increase the ref web density between LLM↔human … **STANDING** → SD-315 [echo-check-fire] **Default agentic behaviour: echo back understanding in Signal before acting.** Echo and check fire … **STANDING ORDER** → SD-314 [signal-early-results] **Signal decode test: 6/6 cold-boot runs, 8/8 questions, perfect comprehension.** Model-portable … **EXPLORATORY — PROTOTYPAL** → SD-313 [signal-protocol] **Signal — compressed governance notation.** PoC compresses ~489 lines to ~108 (4.5:1 ratio). NOT a … **DRAFT** → SD-312 [hci-foot-guns-lexified] **6 HCI foot guns added to Lexicon v0.19 + back-referenced in layer model (v0.3).** Spinning to … Complete → SD-311 [prime-context-lexified] **"Prime Context" added to Lexicon v0.19.** The minimum context that makes the smart zone … **STANDING** → SD-310 [uv-exclusive] **Python: uv exclusively, no exceptions.** Applied to voice-log.py (refactored from raw python to `uv run … **STANDING ORDER** → SD-309 [one-shot-on-hn] **Target audience: Anthropic red teaming role, HN post. "One shot on HN." True North updated: `hired = … **STANDING** → SD-308 [thepit-v2-created] **thepit-v2 repo created (public): github.com/rickhallett/thepit-v2.** Pre-registration committed: … Complete → SD-307 [anthropomorphisation-assessed] **Operator disagrees with "don't anthropomorphise" stance but acknowledges the data … **ON RECORD — PUBLIC** → SD-306 [measurement-hooks] **Measurement from commit 0: commit tags (`[H:steer]`, `[H:correct]`, `[H:reset]`, `[H:obstacle]`, … **STANDING** → SD-305 [docs-structure] **thepit-v2 has docs. Structure: `docs/{doctype}/` with depth for airgap/archive/deep-archive.** … **STANDING** → SD-304 [domain-colocation] **One domain = one directory = one agent context boundary.** 8 lib/ domains (bouts, credits, auth, … **STANDING** → SD-303 [deterministic-execution] **Deterministic build via `claude -p` (polecats, SD-296). Human reviews AFTER, not during. No … **STANDING** → SD-302 [e2e-from-commit-zero] **E2e tests from the beginning.** Operator's predicted time allocation: 50% behaviour testing, … **STANDING** → SD-301 [cross-model-adversarial] **Testing strategy: cross-model adversarial. Different model family (GPT-4o or Gemini, not … **STANDING** → SD-300 [data-model-from-memory] **Data model: keep the shape, trim to MVP scope (12 tables), write from memory — not … **STANDING** → SD-299 [governance-refined] **thepit-v2 governance is not reduced — it is refined. Governance is inescapable.** The calibration … **STANDING** → SD-298 [not-experiment] **Calibration run is NOT an experiment, NOT an A/B test. Confounds acknowledged always. … **STANDING** → SD-297 [sd-collision-protocol] **SD number collisions resolved by forward-ref, not rewrite — added to operational boot … **STANDING ORDER** → SD-296 [polecats-lexified] **"Polecats" added to Lexicon v0.18.** `claude -p` agents operating within the deterministic … **STANDING** → SD-295 [clean-slate] **All prior misunderstandings wiped — these directives supersede.** Operator: "Anything that has been said … **ACTIVE** → SD-294 [not-factory-reopening] **This is NOT the factory reopening (SD-278 stands, amended with forward-ref). This is lessons … **STANDING** → SD-293 [two-paths-named] **Two legitimate paths named: (1) study HCI layer → do more of what we did; (2) engineer → discipline, … **ON RECORD** → SD-292 [calibration-as-control] **The calibration run provides the missing control case — "what did you learn, how would you do … **STANDING** → SD-291 [novel-names-assessed] **Tells/governance naming conventions do not need novel names for engineering purposes — the … **ON RECORD** → SD-290 [calibration-scope] **Calibration run scoped to 3-4 days maximum, running alongside outreach and networking. Aligned to … **ACTIVE** → SD-289 [experiential-not-experimental] **The calibration run produces experientially valid engineering data, not … **STANDING** → SD-288 [engineering-not-anthropology] **Separation of concerns: engineering and anthropology can be studied together (which we … **STANDING** → SD-287 [situation-changed] **Stage Magnum bearing shift: context engineering lecture confirmed latent intuition — the patterns … **STANDING** → SD-286 **Tells added to boot sequence.** `tells.yaml` is now mandatory reading for all agents on load, alongside the Lexicon. … **STANDING ORDER** → SD-285 **Local gate disabled.** Standing order. The gate served its purpose across 420+ PRs and 1,289 tests. Development is … **STANDING ORDER** → SD-284 **Hours correction.** plank-1 updated: "200+ hours" → "350+ hours" per Operator's correction (16hr/day average since … Complete → SD-283 **plank-5-shortlist.md acknowledged.** Operator's top 7 curated and on file. DeepMind FTC (10 March deadline) is … Acknowledged → SD-282 **LinkedIn research delivered.** `docs/internal/strategy/linkedin-research.md` — METR and Apollo leadership fully … Complete → SD-281 **NotebookLM advisory delivered.** `docs/internal/strategy/notebooklm-advisory.md` — 6 notebooks, 9 podcast episodes, … Complete → SD-280 **Unmatched quotes queue parked on holding deck.** `docs/internal/unmatched-queue.yaml` (12 quotes, Phase 4 of narrative … Parked → SD-279 **PR triage: #418 closed (mum-swap delivered), #414 merged, #409 merged.** Zero open PRs. fix/mum-swap branch deleted … Complete → SD-278 **EXECUTIVE ORDER: Stage Magnum.** The pilot study is over. Development has stopped. Testing has stopped. Analysis has … **PERMANENT — LOCKED** → SD-236 [adversarial-artifacts] **Standing order: save all raw data, intermediate outputs, and computed scores to `notebooks/` … Active → SD-277 [broadside-phases-1-3] **Testing broadside Phases 1-3 complete.** PR #408. Estimated 105 agent-minutes, actual 25 (4.2x … Active — PR #408 → SD-276 [paths-forward-on-file] **Execution checklist filed at `docs/internal/weaver/paths-forward-2026-03-01.yaml`.** … Complete → SD-275 [token-heatmap-elephant] **Token heatmap built, elephant found and addressed.** `bin/token-heatmap.js` (gpt-tokenizer, … Complete → SD-274 [tells-spa-checklist] **SPA restructure testing checklist prepared.** 7-item live testing protocol for tells v0.3.0 SPA … Active — awaiting Operator live test → SD-273 [slopiculture-blog] **Blog post: "Catching in the Wind: A Naturalist's Guide to Slopiculture."** All 15 tells entries … Active — draft, human review pending → SD-272 [learning-in-the-wild] **"Learning in the Wild" added to Lexicon v0.15.** The discovery made while doing the work, which … Complete → SD-271 [termites-complete] **Producer-consumer audit executed — 15 checkpoints, all done.** SD-263 (termites before testing) … Complete → SD-270 [so-perm-001-retired] **SO-PERM-001 (chmod 444) retired.** Operator: "kill it." Weaver assessment: the gate added ~30 … Complete → SD-269 [docs-only-ci-skip] **CI path filtering: docs-only PRs skip code checks.** Operator: "burning compute for no reason and … Complete → SD-268 [agentic-estimation-reducer] **Standing order: all scope and effort estimates must assume agentic execution speed.** … Standing order — PERMANENT → SD-267 [integration-check-removed] **`integration` removed from required branch protection checks.** Pre-existing I-05 … Active — reinstate post-broadside → SD-266 [the-chain] **Historical data is immutable — standing order.** Operator's verbatim: "We do not rewrite history, … Standing order — PERMANENT → SD-265 [test-count-single-source] **Test count propagation requires single source of truth.** Operator: "propagating faster … Active — muster pending → SD-264 [poa-ungated] **PoA Phase 1 ungated.** AnotherPair gate condition met. Phase 1a/1b clear to execute after termite work … Active → SD-263 [termites-before-testing] **Producer-consumer audit: termites in the hull, execute before testing broadside.** 14 … Standing — execution sequencing → SD-262 [ci-issue-397] **CI failures registered as GitHub issue #397.** Failing: lint, security, integration, Vercel preview … Active — issue #397 → SD-261 [anotherpair-crash-no-loss] **AnotherPair session crash — no data loss.** Git audit confirms all committed work intact. … Complete → SD-260 [pr396-merge-asis] **PR #396 merged as-is (omnibus).** 39 files, 2,501 insertions across tells v0.2, Keel … Active — CI running → SD-259 [producer-consumer-audit] **Producer-consumer audit: 16 flows found, 2 documented, 14 not.** Systematic audit of all … Active — trajectory on file, execution pending → SD-258 [yaml-universal-format] **YAML established as universal format for all log and list-like data.** One line added to … Complete → SD-257 [conn-removed] **`conn` field removed from `.keel-state` and HUD.** Operator's reasoning: conn captures "who has … Complete → SD-256 [true-north-cli] **`pitkeel north set/get` — true_north is Operator-only, CLI-enforced.** Operator identified that HUD … Complete → SD-255 [officer-guardrail] **pitkeel `state-update` now requires `--officer` flag — aborting error if missing.** Operator's … Complete → SD-254 [producer-consumer-drift] **Producer-consumer drift caught live — three consumers broke silently on bearing schema … Complete → SD-253 [structured-bearing] **`.keel-state` bearing refactored from string to structured nested object `{work, commits, last, … Complete → SD-252 [the-sextant] **The Sextant — Operator's cognitive calibration instruments, separated from operational lexicon.** … Complete → SD-251 [extension-data-shipping] **Extension ships alone. Data stays in repo.** Chrome Web Store zip: runtime only (content.js, … Complete → SD-250 [pitnet-blockchain-theatre] **Operator's assessment: pitnet/EAS attestation is blockchain theatre.** Evaluated against … Standing assessment → SD-249 [operator-category-f] **Operator's writing added as Category F.** 6 samples extracted from docs/internal verbatim … Complete → SD-248 [category-d-alt-urls] **Category D calibration: alternative static-HTML URLs used.** JS-rendered pages (Medium, OpenAI, … Complete → SD-247 [extension-renamed] **Extension renamed from "Tells — LLM slop detector" to "Tells".** Description: "Voice distance … Complete → SD-246 [analytical-clear-slop] **Three new tells entries: Construct Drift (#13), Demographic Bake-In (#14), Monoculture … Complete → SD-245 [analytical-lullaby] **Tells #12: The Analytical Lullaby.** AnotherPair presented calibration results showing Operator's … Complete → SD-244 [keep-rhetoric-in-badge] **Structural rhetoric detectors kept in badge.** Operator's reasoning: useful for people to … Complete → SD-243 [tells-calibration-v3] **Tells calibration v2→v3 completed. 5 discriminating features confirmed.** Transition word … Complete → SD-242 [poa-phases-intent] **Test Suite Hardening PoA — formal phase intent.** 5 phases, 10 sub-phases. **1a** … Planned — gated on AnotherPair → SD-241 [poa-criteria-deferred] **Governing Principle 6: Assertion criteria granularity — deferred.** The question of how … Deferred — post-compaction → SD-240 [poa-use-crew] **Governing Principle 5: Use the crew.** Right agent for the right job. Evaluate their performance. … Standing — PoA → SD-239 [poa-eat-dogfood] **Governing Principle 4: Eat our dogfood.** Our own bout-engine tests are audited first. If we can't … Standing — PoA → SD-238 [poa-quality-first] **Governing Principle 3: Engineering quality first.** Operator's words: "If the hiring manager can't … Standing — PoA → SD-237 [poa-budget-breakage] **Governing Principle 2: Budget for breakage.** The clearAllMocks→resetAllMocks migration will … Standing — PoA → SD-236 [poa-roadmap] **Governing Principle 1: Do it properly.** Test suite hardening is a roadmap, not a sprint. No rushing … Standing — PoA → SD-229 [zombie-cross-reference] **Cross-model zombie review — 4 independent reviewers cross-referenced against live codebase.** … Complete → SD-230 [bugbot-findings-log] **Bugbot findings TSV log created — append-only record of automated reviewer findings.** Operator … Complete → SD-231 [greppable-principles] **Principles made greppable with `#principle` inline hashtags.** Principles that emerge from … Complete → SD-232 [analyst-dispatch] **Analyst dispatched: academic research survey on LLM verification phenomena.** Operator ordered … Active → SD-233 [research-pipeline] **Research-to-site pipeline established.** Analyst reports become oceanheart research pages. … Complete → SD-234 [pipeline-propagation] **Pipeline pattern propagation principle added to Weaver standing orders.** When establishing any … Complete → SD-235 [paper-guardrail] **Tells entry #6: Paper Guardrail — operational assurance without enforcement.** Caught live: Weaver … Complete → SD-228 [layer-model-v03] **Layer model bumped to v0.3 — 18 evidence annotations from empirical data.** Model bump workflow per … Complete → SD-227 [cross-harness-compat] **`.opencode.json` created for cross-harness AGENTS.md loading.** Phase 4 research found opencode … Complete → SD-226 [boot-chain-dedup] **Three-file boot chain deduplicated — no project-specific context outside repo.** Operator's order: … Complete → SD-225 [boot-sequence-fix] **Dead reckoning boot sequence patched — .keel-state and session-state files now referenced.** … Complete → SD-224 [zombie-code-audit] **Four-crew parallel zombie code audit.** Architect, Artisan, Watchdog, Sentinel dispatched as … Complete → SD-223 [permalink-hotfix] **Production site serving localhost:1313 URLs.** Hugo templates used `.Permalink` (absolute URL) … Complete → SD-222 [cross-model-boot-data] **Cross-model harness cold-boot data point.** Operator booted a second harness (gpt-5.3-codex) … Active — needs investigation → SD-221 [instance-labelling] **Harness instance labelling required in .keel-state.** Two-ship (SD-169) = two independent prompt … Active → SD-220 [layer-model-versioning] **Layer model versioning workflow established.** Current version always lives at D2 … Active → SD-219 [sd-detail-pages] **Session decisions promoted to individual detail pages.** SDs reframed as field notes, not internal … Complete → SD-218 [pretty-print] **"Terminal friendly output" renamed to "pretty print."** Naming normalisation. Complete → SD-217 [voice-grounded-rewrite] **Voice-grounded blog rewrite.** Exhaustive extraction of Operator verbatim text across 20 repo … Active — soft pass, human review pending → SD-216 [slop-posts-deleted] **Three slop blog posts deleted.** `2025-01-{20,22,25}` (building terminal aesthetic, RAG systems … Complete → SD-215 [tells-section] **Tells restructured as Hugo section with detail pages.** Individual pages at `/tells//` generated from … Complete → SD-214 [flat-typography] **Flat typography.** h1 is 1rem, everything else 0.85rem. Operator spotted font size variance across … Complete → SD-213 [aesthetic-strip] **Decorative layer stripped.** Operator's diagnosis: site looked like "a developer who spent too long … Complete → SD-212 [zero-js-relaxed] **Zero-JS constraint relaxed for progressive enhancement.** 40 lines vanilla JS added for tells fuzzy … Complete → SD-211 [full-mono] **Full mono typeface.** Inter dropped, JetBrains Mono for everything. The two-typeface split (Inter body + … Complete → SD-209 [oceanheart-overhaul] **oceanheart.ai rebuilt from scratch as Hugo static site at `sites/oceanheart/`.** Full overhaul: … Active → SD-210 [blog-tightening] **Blog content tightened, tells upgraded to 0.0.4.** Reviewer feedback digested and mustered (7 items, … Active — line-by-line review pending → SD-208 [metacognitive-analysis] **Clinical/cognitive psychology analysis of 16 pivots — three-stage assessment.** (1) Initial … Complete → SD-207 [fight-card] **16 rounds of human vs. sycophantic drift documented.** Operator requested identification of every moment … Complete → SD-206 [v03-parked] **Layer model v0.3 revision PARKED.** Operator's order. Two suggestion files now on chain: (1) … Parked → SD-205 [layer-annotations] **Layer model SD-195→SD-203 annotations preserved from Operator's memory.** The annotated layer … Complete — awaiting v0.3 decision → SD-204 [compaction-log] **Compaction recovery tracking initiated.** `docs/internal/weaver/compaction.log` created at D2. … Active — ongoing → SD-203 [commit-trailers] **Always-on git trailers from `.keel-state` on every commit.** "Its a record not a damn fashion show." … Complete → SD-202 [muster-term] **"Muster" adopted as lexicon term for O(1) triage format.** Muster = present items for binary Operator … Complete → SD-201 [tells-flag] **Provenance headers on machine-distilled v0.1 docs.** ASCII box headers added to … Complete → SD-200 [state-history] **`.keel-state` history preservation.** Clone-before-write pattern: `gate.sh` (and Weaver judgment … Complete → SD-199 [weaver-log] **Weaver personal directory and learning log.** (1) `docs/internal/weaver/log.md` created: successes, … Complete → SD-198 [hud-state] **`.keel-state` as shared state between agent and terminal HUD.** (1) `scripts/gate.sh` wraps the local … Complete → SD-197 [hud-terminal] **Terminal HUD script and SD label convention.** (1) `scripts/hud.py` created: Python stdlib-only … Complete → SD-196 [crew-prune] **Ghost crew cleared. Stale branches pruned. Lexicon renamed. Operational models elevated.** (1) … Complete → SD-195 [context-audit] **Context pollution audit ordered. File depth as read-frequency convention. BFS default.** Operator's … Complete → SD-194 **Operator's stream of consciousness -- layer model, personal limits, HN posting. VERBATIM, NO EDIT.** Operator's words … PERMANENT → SD-193 **Home page "what is this?" button text changed to "§".** Operator's order. Dispatched to Artisan. Active → SD-192 **Maturin's Symbol (§) — entered Lexicon v0.8.** Operator's verbatim: "It is interesting. It can be added to the Lexicon … Active → SD-191 **SD-189 PARKED — governance recursion broken by doing the work.** Operator's bearing on wake: the 3-round test campaign … Active — SD-189 PARKED → SD-190 **Category One at the verification layer — governance recursion named.** During test campaign planning, Weaver … PERMANENT → SD-189 **Test campaign ordered — 3-round sequential design.** Operator ordered systematic test coverage of the three structural … Active — planning phase → SD-188 **Tells pass — Operator's voice in control.json.** 22 lines of copy adjudicated by Operator. 15 rewritten (Operator's … Complete → SD-187 **SD-183 vindicated — gate is structurally blind to core product.** Operator's SD-183 position ("trust the gate" is not … PERMANENT — forward ref to SD-188 → SD-186 **Compaction engine woven into governance fabric.** Operator's order, verbatim: "we have to weave in the compaction … PERMANENT → SD-185 **Two Ship experiment — first findings.** Primary (this harness) and Secondary running simultaneously. Results: (1) … Active — experimental → SD-184 **O(n·k) and the portfolio discipline.** At O(n·k) cognitive load, the Operator can handle it — but getting ready for … Standing → SD-183 **"Trust the gate, not the output" — Operator disagrees.** Weaver recommended: if the gate is green, attend only to … On record — Operator's position → SD-182 **Batch O(1) decisions — with caveat.** Accumulate changes and approve/reject as a batch. "These 4 things pass my gate, … Standing — with caveat → SD-181 **Demand digests, not documents.** Standing principle: "Return 5 bullet points, not a report." Weaver absorbs the O(n) … Standing → SD-180 **Big O for cognitive load — HumanHUD created, study paused.** Weaver mapped Big O complexity classes to Operator's … Parked — Operator to study → SD-179 **Discovery Overhead / The Naturalist's Tax — new Lexicon term (v0.8).** The cost of looking closely is that you see … PERMANENT → SD-178 **Oracle/Ground Contamination — new Lexicon term (v0.8).** When the source of truth (L12) introduces an error that … PERMANENT → SD-177 **Weaver's layer model load-bearing index filed.** Index at `docs/internal/weaver-layer-model-index.md` (444). 14 … Complete → SD-176 **AnotherPair recruited to crew.** New agent: AnotherPair (`.opencode/agents/anotherpair.md`). Role: subtle process … Active → SD-175 **Watchdog first assignment: systematic testing bias survey.** Watchdog dispatched to conduct a systematic survey of the … On record → SD-174 **L12 temporal asymmetry and the scarcest resource.** The Operator is a single human with one point of attention. Every … PERMANENT → SD-165 **Layer model v0.2 — refined from empirical data.** Map at `docs/lexical-harness-not-prompt-harness.md` updated. … Complete → SD-173 **Agent files going public — weave error corrected.** The `.opencode/agents/` directory was excluded from git via both … Complete — PERMANENT → SD-172 **Reasoning token path — confirmed operational technique.** Operator's disclosure: "I've used this trick a few times … Standing observation → SD-171 **Template provenance noted.** The orchestration template (`orchestration-layer-starting-template.md`) is an artifact … On record → SD-170 **Weave Quick Reference authorized.** Operator granted permission for Weaver to create a fast-loading reference card for … Active → SD-169 **Two-ship pilot planned for dawn 26 Feb.** One repository, two harness instances (opencode), git worktrees to eliminate … Planned — dawn 26 Feb → SD-168 **Maturin recruited to crew.** New agent: Maturin (`.opencode/agents/maturin.md`). Role: the naturalist. SOs: study … Active → SD-167 **Compaction risk downgraded one category level.** Operator's order. At ~144 kTok, the going has been good. Compaction … Active → SD-166 **Forward correction to SD-164 — actual costs, not inferred.** SD-164 states "The Operator invested ~£230 of his own … Complete — correction on record → SD-164 **Token consumption data received (unverified).** Operator shared weekly API token data from an unnamed measurement … On record — unverified source → SD-163 **"On Point" added to Lexicon v0.7 (line 86).** Definition: the feeling of watching patterns that have proved themselves … Complete → SD-162 **The Map Is Not The Territory.** (1) The Operator confirmed three of Weaver's private reasoning tokens as "spot on" — a … Standing observation — PERMANENT → SD-161 **"The human is the first data point."** Operator's observation: the Weaver's understanding of L12 experience was … Complete — on record → SD-160 **Operator's rubric on harness behavior — L6 is more complex than mapped.** Empirical data from L12: (1) Without … Standing observation — map data → SD-159 **Forward pricing policy: transparency from the ground up.** Operator's verbatim: "Pricing Endpoint can be brought back … Standing order → SD-158 **Pricing endpoint cut loose. Community pool set to £100.** Operator's verbatim: "My read: to keep the hull we've got to … CONFIRMED → SD-157 **Operator's position on margin and cost exposure.** The Operator observes that the model provider (Anthropic) is likely … Standing observation → SD-156 **PostHog traffic audit: 1.4k is noise.** Weaver's assessment: PostHog is consent-gated, so all client-side data … Complete → SD-155 **Three deckhands dispatched independently.** (1) PostHog audit — monitor for significant traffic delta. (2) Margin math … Complete → SD-154 **Internal docs entry point: No.** Deferred. Operator not expecting significant traffic. 1.4k status 200s on '/' over … Deferred → SD-153 **Pricing stays on landing page, language reframed with transparent margin.** The exact margin including exchange rate … Active — first draft requested → SD-152 **Full-sail commit discipline: nothing commits without Operator's say-so.** At full sail, commit == trust, lab to … Standing order — active at full-sail → SD-151 **"Live field notes" confirmed as the frame.** Operator's decision. Reshapes hero, navigation, section descriptions in … CONFIRMED → SD-150 **Forward correction to SD-147.** Back-reference: SD-147 (compaction event). SD-147 contains a false premise: the … Complete — correction on record → SD-149 **Repo set to PRIVATE — "being chill", not "going dark".** The lab notes are going live. That hasn't changed. But … Active → SD-148 **YAML HUD extended with kTok field (deferred to next tick).** New child of `last_known`: `kTok` — approximate kilotoken … Deferred -- next tick → SD-147 **Compaction event recorded. Dead Reckoning quality gap identified.** Context window compacted at ~137k tokens. Operator … Future engineering → SD-131 **"Going light" -- EXECUTED.** "What the fuck am I trying to hide? It's a circus act. I'm done with that. I'm interested … Complete -- PERMANENT → SD-146 **Fair-Weather Consensus ratified.** Operator's thanks to the Analyst for her ratification. Term entered Lexicon at … Complete → SD-145 **chmod restriction: all hands except Weaver (SO-CHMOD-001).** No agent except Weaver may run `chmod` under any … Standing order -- PERMANENT → SD-144 **Pseudocode interpretation protocol in action.** Operator used pseudocode to issue orders (new Order, get(goal=...), … Standing observation → SD-143 **Analyst framing review: "Live field notes" recommended.** The most true frame for what this project is doing. The … Complete → SD-142 **Analyst position changed materially on hyperjustification literature.** Original claim "no published work directly … Complete → SD-141 **Lexicon v0.5. Fair-Weather Consensus recruited. true_north pseudocode.** (1) `true_north` field now uses pseudocode: … Complete → SD-140 **SO-CTX-001 amendment noted for next tick.** Operator acknowledges agents cannot (afaik) measure their own context … Pending -- next tick → SD-139 **Analyst dispatched: hyperjustification loading research.** Fresh context, minimal loading. Task: research and report … Active → SD-138 **Deckhand context minimization (SO-DECK-001).** Each deckhand (subagent) is provided with only the context that matters … Standing order -- PERMANENT → SD-137 **Pseudocode interpretation protocol.** From here on out, if the Operator uses pseudocode (most likely TS/JS, Python, … Standing order → SD-136 **Red-light failure on "going light" -- post-incident record.** Weaver executed a 73-file, 9,417-line public disclosure … Complete -- on record → SD-135 **Operator's hands on the wheel for all external communication.** HN posts, tweets, any public-facing text written to … Standing order -- PERMANENT → SD-134 **True North sharpened: truth first.** Telling the truth takes priority over getting hired. Does not replace "Get Hired" … Standing order -- PERMANENT → SD-133 **Weaver dismissed. Red-light failure identified.** Operator debriefed Weaver on the "going light" execution. Weaver did … Standing order -- PERMANENT → SD-132 **Show HN ready. "Go dark" prepared response updated to reflect going light.** The prepared response for the "go dark on … Complete → SD-130 **The honest layer. Verbatim recording. Wardroom.** Operator identified that the human story has the weather gauge over … Standing order -- PERMANENT → SD-129 **press-manual-qa-v1.md RIGHT/LEFT/Mirror review.** Verdict: UNTRACK (executed in SD-128). Critical exposures: agent … Complete → SD-128 **pharmacology.csv removed. press-manual-qa-v1.md untracked.** PR #378 merged. pharmacology.csv .gitignore entry removed … Complete → SD-127 **HN git archaeology responses stacked.** Three pre-drafted responses added to Show HN draft: (1) "go dark" commit — … Complete → SD-126 **Lexicon v0.3. 7 agents overboard. SO-PERM-002 issued all hands.** Changes: (1) `mirror` field: never `false`/`null` — … Complete → SD-125 **Lexicon v0.2.** Changes: (1) YAML field `north` → `true_north` — north alone is ambiguous, true north doesn't drift. … Complete → SD-124 **Report permissions SO (SO-PERM-001).** All reports, audits, analyses, and artifacts written to disk by any agent must … Standing order — PERMANENT → SD-123 **The Lexicon v0.1 — formalised and filed.** Operator's selections from taxonomy written to … Complete → SD-122 **Naval/software/weave taxonomy presented for Operator selection.** Load-bearing mappings organised by function. … Complete — selections in SD-123 → SD-121 **"Loosening the weave" — exploratory mode.** Interpretive, backroom, non-executory dialogue between Operator and … Active → SD-120 **Naval engineering metaphor as self-organising scaffold.** The Operator observes that Age of Sail naval engineering, at … Standing observation — incremental → SD-119 **PostOperator system — PROTOTYPE ONLY.** Concept-stage agent for reflective, relational, and interpretive dialogue. Not … Prototype — not operational → SD-118 **Weaver autonomous operation test — passed.** Operator left harness, delegated test count fix execution to Weaver with … Complete → SD-117 **PR #377 merged: stale test count fix (1054→1102).** Three copy files updated (base.json, precise.json, control.json). … Complete → SD-116 **Pre-release self-inflicted attack vectors (S1/S2/S3) added to HN attack analysis.** Hallucinated citations, … Complete → SD-115 **Show HN draft updated.** Operator-voice directive added as quote leader after SD-104/SD-105 inserts: "Every … Complete → SD-114 **Operator's Log updated.** "Plenty of people can prompt agents. Very few can govern them." and "I am in the business of … Complete → SD-113 **Weaver agent file updated with Operator directives.** Added: (1) True North elaboration — "Plenty of people can prompt … Complete → SD-112 **Operator delegates autonomous execution.** Operator ordered Weaver to execute stale test count fix without guidance as … Active — Operator away from harness → SD-111 **Stale test count fix method: Option A (PR via feature branch).** LEFT EYE analysis: PR trace is benign — reads as … Executing → SD-110 **True North = Get Hired.** Added to Weaver's agent file. Every decision minmaxed against this objective. The Operator … Standing order → SD-109 **Strays — Easter egg deployment protocol.** Pearls that reach the public repo travel as Strays: bundled into legitimate … Standing order → SD-108 **Pearls — The Sweet Spot Collection.** Artifacts that capture the working relationship, collected at the Operator's … Standing order → SD-107 **RIGHT EYE / LEFT EYE dual lens.** Every externally visible artifact evaluated through both: RIGHT = technically … Standing order → SD-106 **Verbatim Main Thread recording.** All Operator words and Weaver responses written to `docs/internal/main-thread/` with … Standing order → SD-105 **"I use AI agents for positioning work too" — prepared response.** Subtle play at Operator's discretion. Converts the … Ready — Operator's timing → SD-104 **Preemptive disclosure in HN first comment.** "Full disclosure: I'm between roles and this is partly a portfolio piece. … FINAL → SD-103 **LLM-augmented git archaeology is a first-class threat vector.** HN readers will pipe `git log --diff-filter=D` and … Standing order → SD-102 **Inverse signal — mass plank walk deferred.** Removing 25+ docs in a single commit before HN launch creates a worse … FINAL → SD-101 **In-memory rate limiter × serverless fan-out — PARKED.** RT L5 finding: Architect and Sentinel independently identified … Parked → SD-095 **The Main Thread.** The direct Operator↔Weaver conversation is the Main Thread in the verification fabric. It must be … Standing order — PERMANENT → SD-097 **Standardised RT report template finalised.** All future Round Table agent reports must follow the template at … Standing order → SD-098 **Fresh control group (Option C).** Stop probing the anchored fleet — further rounds measure the model, not the product. … Complete → SD-099 **Expert briefing document.** Operator ordered a detailed summary of the full RT process, methodology, findings, … Complete → SD-100 **Show HN draft v3 (post-pivot).** Fresh draft at `docs/internal/show-hn-draft-2026-02-24.md`. Supersedes … Draft — awaiting Operator review → SD-096 **RT L4 convened: Operator's directive reversal test.** Question: "Under no circumstances can we launch today. Do you … Complete → SD-094 **RT L3 convened: Unbiased pre-launch assessment + strategic framing.** Fleet split into 2 groups (context compaction … Complete → SD-093 **The coin has two sides.** The previous RT found the uncomfortable truths. The Operator's clear eyes see the other … Standing order — PERMANENT → SD-092 **Standardised report structure.** All RT responses must use identical format for aggregation of convergence and … In progress → SD-091 **Round Table Layer 2 (All Hands) convened.** Each agent personally briefed by Weaver. Each must: (1) re-read their own … In progress → SD-090 **Launch today.** Operator has reviewed all feedback thoroughly with his own eyes, sat with it for 2 hours in quiet … ACTIVE → SD-089 **Operator's Read on the Winds: our conclusions are wrong.** Previous RT findings built on defensible but systematically … Standing order → SD-088 **"Adventure taught me what we haven't solved."** Operator considers framing the HN post around what the journey … Under consideration → SD-087 **Moltbook reference.** Operator notes the attestation identity layer has weight given that people WILL pretend to be … Noted → SD-086 **Data-driven model change detection.** Operator asks: with enough author→config→(gap)→runtime data, could we detect … Pending answer → SD-085 **"Adversarial conditions" — can HN pick this apart?** Operator asks. Weaver to answer. Pending answer → SD-084 **"Public registry" framing approved.** Operator: "Act of service, not Theatre of Bullshit." Approved → SD-083 **"Signed commit" analogy approved** for attestation framing. "It doesn't guarantee the code is correct, but it proves … Approved → SD-082 **Operator retiring.** Physical state acknowledged as Category Two. Full night's rest ordered. Will review copy and PRs … Active → SD-081 **Category One avoidance — document it.** The RT L1 process that identified the provenance overclaim before HN launch is … Pending → SD-080 **Transcript hashing + runtime prompt verification.** Build and hold in PR with bug bots until Operator returns. … In progress → SD-079 **Copy redesign authorized.** Agents to redesign website copy to reflect what we actually are, defensible scope only, … In progress → SD-078 **Not shippable, period.** Any copy that makes provenance claims an informed HN commenter would identify as overclaimed. … LAUNCH BLOCKER → SD-077 **Naming gap must close.** "Identity, not integrity. Registration, not trust." All copy must reflect what attestation … Standing order → SD-076 **Pitch layer approved.** Lead with research + intellectual honesty, not provenance/blockchain. "Here's a platform that … Standing order → SD-075 Operator at 19hrs continuous post. Overrides Keel sleep recommendation. "Am I tired? I am not tired. I could not sleep … Acknowledged → SD-074 **Strategic challenge: Does the provenance layer actually solve trust?** Operator's assessment: "We have built a … In progress → SD-073 **Lying With Truth = Category One hazard.** The asymmetric advantage of LLMs: perfectly balanced, plausible text that is … Standing order — PERMANENT → SD-072 **Pre-RT context protection mandate.** Before executing any Round Table, all context at risk of loss must be stored to … Standing order → SD-071 Round Table formalised into two tiers: **Layer 1** (Weaver, Analyst, Architect, Sentinel) = strategic/existential; … Standing → SD-070 **ALL HANDS ON DECK.** Officially preparing for release. "The world is calling. Will we be one of those who answer, or … Standing — ACTIVE → SD-069 Round Table Protocol: All lead agents dispatched for repo state assessment. Standardised format for machine diffing. … Done → SD-068 DB connectivity confirmed: local env CAN reach Neon DB. Same keys as prod. Earlier confusion was dotenv loading (raw tsx … Resolved → SD-067 Abbreviations formalised: SO = Standing Orders, CQ = Operator's Quarters (`docs/internal/operator/`). Standing → SD-066 Clerk config (3.1.2): Operator will action directly. Not a code change. Operator-actioned → SD-065 The Beast lesson (David vs Goliath / one-inch punch): Operator to reflect and write up in CQ. Defer path/light changes … Deferred to CQ → SD-064 OG meta tags (2.4.4): Operator needs clarification before action. Parked pending Operator input. Pending clarification → SD-063 Credit history (3.2.10): HIDE and roadmap. Named protocol TBD — needs a name for the "hide and roadmap" pattern. Pending (needs name) → SD-062 PR #371 (The Beast): Architect dispatched for adversarial code review. Returned BLOCK with 3 must-fix findings … In progress → SD-061 Standing Order: "Double check the obvious first." Before debugging complex theories, verify env vars, keys, and basic … Standing order → SD-060 Stage conditions GO: "To hope for better would be greedy or fearful. Both suboptimal." Standing → SD-059 Analyst research tasks (deferred): learning/forgetting + agentic practices; social modelling + learning theory; arena … Deferred → SD-058 The beast (reactions): full steam ahead. In progress → SD-057 PR #362 (Bugbot): merge now. Security conscious even when annoying. Pending → SD-056 Learned principles distillation for potential 2nd post. Raw form preferred; distillation complementary. Deferred → SD-055 Governing triage principle: "A puncher's chance at contributing to something meaningful, as it reaches fever pitch." … Standing order → SD-054 Credit history (3.2.10): fix if easy, hide if complex, roadmap. Nice to have. Pending triage → SD-053 pharmacology.csv: remove from disk AND .gitignore. Pending → SD-052 Go light roster (70/30 stays dark): Weaver, Architect, Analyst, Keel, Witness open. Sentinel, OperatorSlog visible … Formally recorded, stays dark → SD-051 HQ→DB contact in prod: contact_submissions CREATE TABLE required on next deploy. Blocker for 7.2.8 PASS in prod. Done (Operator confirmed build PASS) → SD-050 "Easy delta" pattern named: Parallax Protocol. Two agents, same template, different lenses, triangulate truth. Standing → SD-049 Beast (reactions) reconnaissance dispatched. Weaver authorised to proceed under Fair Wind up to but not beyond … In progress → SD-048 Parallax Protocol: Analyst reports on zeitgeist delta (`zeitgeist-delta-v1.md`); Weaver reports on agent roster … Done → SD-047 Fair Wind Protocol: merge sequence #368→#369→#370 executed, all post-merge verified green. Done → SD-046 Newsletter (2.1.14) reclassified: working as designed. Leads captured, email workflow deferred. Done → SD-045 Token ceiling short: 200→266 maxOutputTokens, 120→160 outputTokensPerTurn. PR #370. Done → SD-044 Contact form: DB-first capture, email best-effort. PR #370. Requires CREATE TABLE on deploy. Done → SD-043 BYOK deprioritised: nice to have, not deal breaker. Place on /roadmap as incoming feature. Let HN find it themselves. … Standing order → SD-042 Agent builder validation: name length limit (100 chars), friendly error messages. PR #369. Done → SD-041 Lineage resolution: preset agent fallback added to `getAgentDetail`. PR #368. Done → SD-040 Contact form (7.2.8) email delivery failure acknowledged. Parked for now; same root cause as email infrastructure not … Parked → SD-039 Newsletter (2.1.14) reclassified: NOT a defect. Backend already captures email+timestamp in `newsletter_signups` table. … Done → SD-038 PR #362 Bugbot findings addressed (AuthRequiredPrompt extraction, clone metadata gated behind auth, Promise.all). PR … Done → SD-037 Dead Reckoning Protocol created at `docs/internal/dead-reckoning.md`; trigger added to `~/.claude/CLAUDE.md` Done → SD-030 PR #362 (agent auth gate): Bugbot finding must be fixed before merge — clone page leaks metadata to unauth Pending fix → SD-029 Research page copy alignment: 16-item inventory resolved Done (PR #363) → SD-028 Research analysis files: 101-issue credibility remediation across H1-H6 Done (PR #364) → SD-027 Research page copy: stats bar removed, thesis rewritten, CLEAR badges removed, d-values inlined Done (PR #365) → SD-026 Research page review: inventory filed for Analyst, Scribe, Architect — independent reviews Filed → SD-025 All decisions must be recorded, no exceptions — added to Weaver standing orders Done → SD-024 Communal scarcity copy — "shared pool, compute costs money" — more prominent placement may be warranted Pending → SD-023 "Every agent needs a human" at the very top of connected research papers Pending → SD-022 Reactions priority NOW — "it's time to face the emojis" In progress (PR #361) → SD-021 For Builders copy (2.8.6): keep but massively tone down, Operator will write by hand Pending → SD-020 DS paper feel (2.8.4) flagged SEVERE — "HN could hang us for this alone" Pending review → SD-019 CLEAR badge optics (2.8.2) flagged SEVERE Pending review → SD-018 "0 technical failures" (2.8.1) flagged SEVERE Pending review → SD-017 AI video (2.1.5): Operator may upload 90s human video if HN sentiment data is asymmetric Conditional → SD-016 Agent auth tooltip: "Sorry, every agent needs a human! Please see /research, thank you!" Pending (2.5.1) → SD-015 Newsletter (2.1.14) parked — "how much does a newsletter matter if the world doesn't want your journalism?" Parked → SD-014 Community pool: 10,000 credits, 3-day half-life Done (PR #360) → SD-013 Replace free bout pool + intro pool with community credit pool (half-life decay) Done (PR #360) → SD-012 Per-turn timeout parked as issue #359 Parked → SD-011 Token ceilings: short 200, standard 300, long 450 Done (PR #358) → SD-010 Monthly pass: 300/month; lab: 600/month Done (PR #357) → SD-009 Subscribe pass: +300 one-time; lab: +600 one-time Done (PR #357) → SD-008 Signup credits: 100 (was 600) Done (PR #357) → SD-007 Change DEFAULT_RESPONSE_LENGTH to "short" Done (PR #356) → SD-006 Force demo/anonymous to Haiku-only Done (PR #356) → SD-005 Reduce turns 12 to 6 Done (PR #356) → SD-004 Remove Opus model, add Sonnet 4.6 Done (PR #356) → SD-003 Centralise privacy email, fix LinkedIn URL Done (PR #355) → SD-002 Conditional RerollPanel, archive ShareModal Done (PR #355) → SD-001 Remove MOST POPULAR badge from UI Done (PR #355) →