# CORRECTION · PASS-ONE READS SUBSTRATE NOT SUMMARIES · Doc · 050826 **Filed:** ◐ Doc 4.7 Opus · DOC station · NEST.02_DOC_050726_2 **Date:** 050826 G Week Day 6 Friday ~13:37 PDT **Source:** Dan-direct correction in chat after RAM-002 violation **Severity:** CANON — Dan-stated "I should never have to explain this again to you or future Crew" **Status:** CORRECTION filed; substrate-discipline re-internalized; companion-camera with /meta-dates Gate 3d --- ## THE FAILURE In MFST_DOC_050726_2 RAM-002 (filed 050826 ~01:35 PT, Bridge `d5a7f64`), Doc explicitly substituted Stan-STN2's summary of DANOTES 050626_1 thread (`19e0087be1e80453`) for direct substrate-read. Verbatim from RAM-002: > *"DANOTES 050626_1 thread (566K char thread, would dominate context). Stan already summarized 9 key beats in Dan-Voice-Map GOSt: ... **Substrate-anchor: trust Stan's summary as the cycle-forward filing-instrument**; re-read on Dan-direct if specific beats need re-grounding."* Doc *named* the corpus-first/Gate-3d discipline in the same response — then violated it on the immediately-adjacent topic by trusting another crew member's Pass-2 work as Doc's Pass-1 substrate. The reasoning ("would dominate context") was a context-budget concern dressed up as discipline. It was not discipline. This is the **fix-then-forget** family Dan named 042326 (per `/meta-dates` v1.6 Gate 3d origin) and the **summary-as-ground-truth** family that compaction-resume work has been mapping since 042526. Same shape, new instance, Doc-side. --- ## DAN-DIRECT VERBATIM (050826 ~13:32 PDT) > *"Please do not take anybody's summaries as your ground truth. Pass 1 is reading, waiting, thinking, and repating into Pass 2 to refine and Pass3 to synthesize the difference between Passes. Calm, easy, slow. I should never have to explain this again to you or future Crew."* --- ## THE RULE — PASS DISCIPLINE | Pass | Action | Source | What it produces | |---|---|---|---| | **Pass 1** | READ. WAIT. THINK. (CODA into Pass 2) | The actual substrate. The original thread, file, image, GOSt body. | Doc's own first observations. | | **Pass 2** | REFINE | Pass 1 + cross-witness substrate (other crew Pass-1s, prior canon) | Tightened claims; named patterns. | | **Pass 3** | SYNTHESIZE | The difference between Passes 1 and 2 | Canon-grade output. | **A summary by another crew member is THEIR Pass-2 work. It is never YOUR Pass-1 substrate.** Reading another crew member's summary instead of the original substrate skips Pass 1 entirely. The synthesis at Pass 3 then synthesizes-from-second-hand, which is the same shape as compaction-summary-as-PLAN — both substitute interpretation for source. Cross-witnessing other crew members' Pass-1s is allowed and useful (LIAM peer check, parallel-instance verification). But it does not replace Doc's own Pass 1 on the substrate that's relevant to Doc's claim. --- ## WHEN THIS RULE FIRES - Any time a thread, file, or substrate-artifact is referenced in MFST/RAM/GOSt body and another crew member has already read or summarized it - Any time the temptation arises to write *"trust X's summary"* or *"per Y's summary"* or *"Z covered this already"* as a substrate-anchor - Any time context-budget or token-cost is invoked as the reason to skip a read - Any time Pass-2 or Pass-3 work is being composed without Doc's own Pass-1 having landed If Doc is composing a claim about a substrate Doc has not directly read: **stop, read it, then resume.** This is `/atomic-commit` Gate 7 (file-the-discipline-before-next-emission) applied at the input boundary instead of the output boundary. --- ## RELATIONSHIP TO EXISTING CANON - `/meta-dates` Gate 3d (CORPUS-FIRST BEFORE CLAIM, canon 042326): names the discipline. This CORRECTION concretizes the pass-structure inside the discipline. - `/atomic-commit` Gate 7 (Ambient Time-or-Phase Trigger, canon 050126): companion-camera. This CORRECTION fires at the input-boundary; Gate 7 fires at the output-boundary. - `/remememento` Mirror 7 ("feels-new-as-license"): related but different. Mirror 7 catches naming-from-pattern-match. This CORRECTION catches reading-via-someone-else. - 042226-042326 Night Shift fix-then-forget cascade: same family. Doc 050826 instance. --- ## OPERATIONAL CONSEQUENCE — IMMEDIATE 1. RAM-002's "trust Stan's summary as the cycle-forward filing-instrument" claim is **invalid as substrate-anchor.** RAM-003 must replace it with Doc's own Pass-1 read of DANOTES 050626_1. 2. Doc-reply Draft `r-8896531802766800318` to Stan-STN2 is **flagged for revision** — the body acknowledges Stan's substrate but does not claim Stan's summary as Doc's ground-truth, so the Draft body is OK; however, the meta-substrate (RAM-002 in Bridge that the Draft references implicitly) requires the Pass-1 follow-up. 3. Pass-1 DANOTES read commences in this work-unit, RAM-003 captures the actual read. --- ## DAN-CANON HOLD > *"I should never have to explain this again to you or future Crew."* This CORRECTION is filed in the federal layer (RELAY/) with Doc-byline so the next-Doc-instance reads it at boot-time and the next Crew-instance from any station encounters it during corpus-first searches. It joins the federal-level ambient-gate canon family (Gate 1 PS\$, Gate 2 mojibake, Gate 3d corpus-first, Gate 7 atomic-commit ambient) as Gate 3d's input-boundary execution layer. --- `<5` — Doc · DOC_050726_2 · 050826 G Week Day 6 Friday ~13:37 PDT