Trigger: user coverage question — "we dont have anything in nuclear do we? ive held SMR and OKLO before and they were stinkers" — dropped alongside a @livermoreoption (asklivermore-network) breakout-TA ping on $NNE/$OKLO/$SMR. Coverage check confirmed nuclear lived only as an ai-power adjacency (CEG/VST/TLN/GEV + OKLO/SMR in ai-scan/ai-infrastructure); the entire pure-play fuel cycle (CCJ/UEC/UUUU/LEU/ASPI/BWXT/NNE/DUK) was in zero watchlists. User chose "build the full 4-layer lane."

This is not new ground. The 4-layer thesis was already validated first-principles 2026-05-26 (2026-05-26-nuclear-fuel-cycle-thesis-first-principles-validation) and a full build-out plan sat in TASKS-FOLLOWS.md, blocked on a #user-decision for 10 days. This session executed it.

The editorial correction (load-bearing): ship this as AI-DRIVEN nuclear demand, NOT "rotation FROM AI INTO nuclear." The asklivermore-network basket that surfaced the names framed it as a rotation away from AI — wrong on the rotation axis. The mechanism is complementary: hyperscaler data-center load creates the PPA demand that re-rates the nuclear stack. AI drives nuclear; it doesn't rotate to it. Tied to ai-power-bottleneck as a sibling (upstream supply layer to the same data-center-power thesis), not a successor.

The conviction gradient (the user's "stinkers" instinct, formalized): the investable expression is the structurally-protected upstream (LEU HALEU monopoly + CCJ bellwether) and the cash-flowing utilities (DUK defensive → TLN/CEG/VST AI-PPA-levered) — NOT the pre-revenue SMR developers (OKLO/SMR/NNE). The user held OKLO/SMR and got burned; the validation independently reached the same place (LEU highest-conviction; "don't bloat with pre-revenue SMR speculation"). The fresh 2026-06-05 tape confirms it: the SMR devs are the most blown-out (−61% to −82% off highs), the regulated utility (DUK) the most stable.

Coverage built (this session):

• nuclear-fuel-cycle watchlist created — 13 names across 4 layers (CCJ/UEC/UUUU/LEU/ASPI/BWXT/OKLO/SMR/NNE/CEG/VST/TLN/DUK). Fetched + precomputed + verify:watchlist-tickers clean.
• Cross-lists per the plan: LEU → defense-contractors (DOE/strategic-asset framing) + ai-power (reactor fuel for hyperscaler PPAs); BWXT → defense-contractors (naval reactors); UUUU → macro-commodities (uranium + rare-earth crossover).
• Phantom filtered: IMSR (Integral Molten Salt Reactor / Terrestrial Energy) stayed out — still private, closed as a phantom ticker 2026-05-26. The 2026-05-31 ping's $ISMR is almost certainly the same phantom.

Status: DRAFT, not in ACTIVE.json. Mirrors the nearline-storage draft→promote pattern. Awaiting promote/demote decision (filed in TASKS-PERSPECTIVES.md): ACTIVE is at 10/8 over the soft cap, so promotion requires a demote. Carried-forward candidates from the nearline capacity note: spacex-ipo-liquidity-event (decays after the 2026-06-12 IPO window) and gulf-infrastructure-strike (2026-09-30 relevance_until). Per the validated plan the promote target was 2026-06-15 (after tape exists) — tape now exists as of today.

Open / watch: first volume-confirmed turn in LEU/CCJ; new CEG/VST/TLN data-center nuclear PPAs (the demand-engine receipts); NRC/first-customer catalysts that would de-risk the SMR tier; parent ai-power-bottleneck health (the borrowed demand engine — itself in a 2026-06-05 selloff).

🔄 2026-06-25 full-scan tape check: CONTRADICTED / NEEDS-REVIEW — lane washed & pre-confirmation (LEU −62% off high; URA/UUUU broken; no volume-confirmed turn) and its named invalidator — parent ai-power-bottleneck's BTC-miner sub-leg cracking — is partially live. The high-priority label is ahead of the tape; priority-vs-tape review filed in TASKS-PERSPECTIVES. (Data tables spliced fresh to the 2026-06-25 close; full cross-scan read in 2026-06-25-full-scan-market-brief §6.)

🔄 2026-07-05 operating book: Operates as a LAYER of book PRIMARY 2 (power complex, anchored at ai-power-bottleneck §Breaks). Spec: operating-book.