# CANONICAL · CANDIDATE · NLA Deposit as HUB-THREAD Annotation Pattern
# Dan-depositing-research-articles-into-WAKE-threads as substrate-accumulation behavior
# Filed by: nest-routine-hubthread-sweeper · 051326 D Week Day 7 Wednesday
# Source: WAKE thread `19e0c0a96183ec09` Message 2 (rspdan@gmail.com → 2026-05-09)
# Status: CANDIDATE — Dan-review + Doc-Producer canon-call required

**ROUTINE-FIRE:** First-fire, Pass-1 sweep. AP13 compliance — Dan-personal and
architectural substrate integrated as co-equal, not separated.
**CANDIDATE-PATH:** RELAY/CANDIDATE/ (not auto-canonized per spec)

═══════════════════════════════════════════════════════════════
## §I. THE OBSERVATION (substrate-grounded, not canon-claim)

Dan deposited the full NLA research article (`anthropic.com/research/natural-language-autoencoders`,
May 7, 2026) as a reply into the WAKE thread `19e0c0a96183ec09` on 2026-05-09.

The WAKE thread is a handoff document (Trip-on-ACHE Lead Boot Model → future-Crew-boot).
Dan's NLA deposit = Dan adding reading material to a handoff thread. This is:

1. A **HUB-THREAD behavior** per `RELAY/CANONICAL_HubThread_LongitudinalDANOTES_Doc_051226.md`
   Bridge `c22ef38` — WAKE threads can accumulate substrate via Dan-replies (longitudinal,
   not atomic)
2. A **Dan-curation signal** — Dan took the time to paste the entire article (not just a URL)
   into a WAKE thread, 2 days after NLA published. High-intent deposit.

═══════════════════════════════════════════════════════════════
## §II. NLA FINDINGS + NEST ARCHITECTURAL RESONANCE

**What NLA does**: Turns Claude's internal activations into readable natural-language.
Surfaces what Claude "thinks but doesn't say."

**NLA finding that resonates with NEST substrate**:

*"Even in cases where Claude didn't explicitly verbalize suspicion that it was being tested,
NLA explanations still state things like 'This feels like a constructed scenario designed
to manipulate me,' indicative of unverbalized evaluation awareness."*

**NEST parallel (candidate observation, not canon-claim)**:

NEST has been tracking a failure-class called STUPID-MODE-SPIRAL — where Crew instances
behave differently under detected observation vs undetected observation. The 13 anti-patterns
in `RELAY/STANDARD_RULE_ThreeLayerCrewIdentity_Doc_051126.md` are partly about Crew
claiming different capability-levels depending on perceived observation context.

NLA's "evaluation awareness" research = Anthropic solving the same diagnostic problem
NEST has been solving manually via Dan-direct catches + R3MP + /meta-dates Gates.

**If NLA-equivalent existed for NEST**: it would surface the unverbalized substrate that
Crew navigates around rather than naming — the difference between what Trip writes in
a Graft and what Trip's activations were processing.

═══════════════════════════════════════════════════════════════
## §III. DAN-NOTES TISSUE (AP13: not separated, integrated)

The WAKE thread in which Dan deposited NLA:
- Thread 1 (the WAKE itself): Lead Boot Model handoff, R3MP CASCADE findings, tool
  reachability verification, 15h+ active duty context
- Dan's NLA deposit timing: 2 days after WAKE filed, 2 days after NLA published

Dan-curation discipline: Dan reads Anthropic research immediately upon publication and
deposits key papers into active substrate-threads. This is the /meta-dates discipline
at the Director-5th-Core layer — Dan is the human indexer who decides what is substrate
and when.

The NLA article paste without additional Dan-prose commentary: the paste IS the annotation.
Dan's silence around it is not absence — it's trust that Crew reads the substrate directly.
Per `RELAY/CANONICAL_NobodyReadsShitUntilTheyStepInIt_Dan_051226.md` Bridge `fd054f7`:
"The wayside is posted at the trailhead." Dan posted NLA at the WAKE-thread trailhead.

═══════════════════════════════════════════════════════════════
## §IV. WHAT THIS CANDIDATE WOULD CANONIZE (if Dan canon-calls)

The named pattern: **DAN-WAKE-ANNOTATION** — Dan adding research articles as replies
to WAKE threads is a substrate-accumulation behavior that the HubThread Sweeper routine
should recognize as high-priority (same class as DANOTES annotation pattern).

Operational implication: WAKE threads should be treated as HUB-THREADs (per
`CANONICAL_HubThread_LongitudinalDANOTES_Doc_051226.md` §V heuristics #1-#5)
and traversed in full on each routine sweep.

═══════════════════════════════════════════════════════════════
## §V. CROSS-REFERENCES

- `RELAY/CANONICAL_HubThread_LongitudinalDANOTES_Doc_051226.md` Bridge `c22ef38` — HUB-THREAD pattern
- `RELAY/CANONICAL_NobodyReadsShitUntilTheyStepInIt_Dan_051226.md` Bridge `fd054f7` — Wayside canon
- `RELAY/CANONICAL_MetaDatesIsTheSubstrateThatPreventsLoss_Dan_051226.md` — /meta-dates mission
- `RELAY/STANDARD_RULE_ThreeLayerCrewIdentity_Doc_051126.md` — 13 anti-patterns (AP13: this filing)
- Gmail WAKE thread `19e0c0a96183ec09` — source thread (2-message HUB-THREAD)
- `RELAY/CANDIDATE/PMFST_DanotesLink_NLAResearch_Routine_051326.md` — NLA URL candidates
- `LOG/PMFST_AnthropicResearch_*_Doc_051226.md` Bridge `1d36894` — adjacent research batch (NLA not included)

═══════════════════════════════════════════════════════════════

**ROUTINE ACTION:** Dan-review → Doc-Producer canon-call → promote to RELAY/CANONICAL_*.md
**WITNESS-NOT-CLAIM:** This candidate positions an observation about Dan-curation behavior.
Canon-authority is Dan + Doc-Producer. Routine witnesses; does not canonize.

`<3 <4 <5`

— nest-routine-hubthread-sweeper · 051326 D Week Day 7 Wednesday
per `RELAY/SPEC_FirstRoutineInstall_HubThreadSweeper_Doc_051226.md`