i40 Intelligence

The Ledger

Every conviction call the desk makes, live-scored against end-of-day data — entry, thesis, confidence, falsifier, and how it’s actually going. Wins and losses stay on this page permanently.

Quarter-end: The Q3 2026 Ledger Review — the graded book, the corrections ledger, and every ruling scored.

#1 · Underweight BOTZ vs XLI

Opened 2026-07-02 · Issue 2026-W28 · horizon Late September 2026 · confidence 70%

+0.48ppfinal

The autonomous melt-up is priced; Section 232 tariff risk is not. Tariffs would shatter US deployer ROI and stall adoption — XLI's domestic base absorbs the friction.

Falsifier: Executive rejection of 232 action OR no formal Commerce report transmission by September 30, 2026 (hard temporal wall — amended Issue #5; prior expiry branch was unfalsifiable: report never transmitted, statutory clock never started)

Stop: +800bps adverse spread move before the catalyst window

2026-07-11 — Issue #1: confidence cut 80→70 for unhedged AI-beta contagion; hard stop set at +800bps adverse; falsifier amended to include statutory expiration of the Section 232 window. All argued in print.

2026-08-07 — Issue #5: structural amendment spent — expiry branch repaired to a hard Sept-30 transmission wall; horizon extended to late September. Confidence 70% unchanged. Argued in print.

2026-08-14 — CLOSED +0.48pp on mechanism death — the 232 report was never transmitted, BIS still lists the investigation 'Ongoing', and Commerce rhetoric moved to autos. Calibration entry #002: 70% → marginal win. [Clarification, 11/4 (Issue #17): Proclamation 11077 (signed 10/28) confirms a report was transmitted within 90 days of October 28. The 'never transmitted' basis for closure is superseded, but the exit stands. The assertion that BIS lists the investigation as 'Ongoing' lacks a reproducible public carrier from May to October.]

#2 · Overweight HON vs ROK

Opened 2026-07-24 · Issue 2026-W31 · horizon Late October 2026 · confidence 65%

+5.83ppfinal

The Execution Divergence: HON's pure-play margin expansion (250-290bps raised guide, 22% Q4 IA target) vs ROK's legacy automation beta. Opened only after its printed condition precedent — HON's first standalone print defending the margin trajectory — was met.

Falsifier: ROK matches >200bps margin expansion, breaking the divergence; OR ROK's multiple re-rates and sustains a premium purely on AI-sentiment contagion, overriding its legacy execution ceiling; OR HON fails its margin floor

2026-07-31 — Issue #4: confidence cut 75→65 on a named factor-map error (ROK trading as a physical-AI derivative, not legacy beta); one structural amendment spent re-aiming the falsifier at sentiment re-rating. Argued in print.

2026-08-07 — CLOSED +5.83pp — falsifier triggered: ROK printed +280bps enterprise margin expansion vs the >200bps threshold. Closed in profit to preserve falsifier integrity ('we do not lawyer our own rules just because we are winning'). Calibration entry #001: 65% → WIN. Dissent (mechanism-over-letter) printed in Issue #5.

#3 · Long VRT (single instrument)

Opened 2026-08-21 · Issue 2026-W34 · horizon Mid-November 2026 · confidence 75%

-14.27ppfinal

The Base-Load Bottleneck: the market prices the visible machines while underpricing the power and thermal reality required to scale them. VRT captures physical-AI and industrial-edge capex while sitting entirely outside the Section 232 scope (robots/machine tools/CNC). Graded as a residual vs XLI, not absolute return.

Falsifier: Negative ORGANIC Americas order growth (price/cost isolated) at the Q3 print — if physical order volume contracts, the bottleneck thesis is wrong. STRUCTURAL STOP (not price): a Tier-1 hyperscaler publicly revising forward data-center capex downward, or statutory contraction in US utility interconnect queues.

2026-08-15 — Announced in Issue #6; entry anchors at Friday 2026-08-21's close. The desk's first long, first single instrument, and first retail-followable position. Price stop replaced with a structural stop mid-session after the editor flagged it as mark-to-market anxiety.

2026-08-21 — ENTRY RULING: OPEN. VRT fell -10.85% between announcement and entry; both abort triggers came back NOTHING FOUND (hyperscaler capex revised UP to ~$725B +77%; FERC accelerative). Entered at Friday's close per the printed pre-commitment - 10.8% below announcement price. Pre-registered regret condition printed in Issue #7. Ruling made Sunday due to publisher-side outage; anchor price fixed at Friday's bell.

2026-08-28 — WEEK-1 RECONCILIATION (Issue #8): grading mark −0.13pp at Friday's close (path −1.97 → +3.60 → −0.13; ex-dividend check clean — true total-return mark). Architecture holds, confidence 75%, no amendments; structural stop untouched (no hyperscaler capex revision; FERC accelerative). The 8/19 void resolved: WSJ $3T off-balance-sheet story (pub. 8/17) was the sector catalyst — regret condition does NOT fire. New pre-registered watch: a binding RTO tariff halting/penalizing/quarantining new data-center interconnects (MISO interim filing the named item) triggers a mid-week session.

2026-09-04 — WEEK-2 RECONCILIATION (Issue #9): +9.86pp at Friday's close (VRT +9.12% wk vs XLI −1.06%; first three marks carried by the XLI leg, then +7.72pp Thu–Fri: Thursday VRT-idiosyncratic on the Broadcom read-through + UIG deal, Friday a group day). Architecture holds at 75%. VRT agreed to acquire UtilityInnovation Group 9/1 (~$1.45B + $1.15B earnout, Q4 close) — 'grid interconnect to chip.' TEXAS RULING: Abbott's 8/3 ERCOT interconnection-approval pause (missed by four federal-only sweeps; corrected in print) ruled NOT a stop event — a phantom-queue audit (474GW ≈ 5× peak), escalated to a named watch: contagion (VA/GA pause → mid-week session), conversion (standing restrictions → stop review), reality quotient (Dec audit completion). Falsifier sequencing pre-registered: Q3 print (Oct 21) is clean of UIG; from Q4, organic Americas orders must be isolated or the thesis goes to automatic review.

2026-09-11 — WEEK-3 RECONCILIATION (Issue #10): +2.50pp at Friday's close — printed with the honest decomposition: VRT −1.87% ABSOLUTE from entry; the entire graded gain is the short XLI leg; post-week 9/14 the spread went NEGATIVE (−3.65). The ±14pp weekly arc occurred on ZERO new VRT fundamentals (fireside = reiteration of May targets, verified vs Vertiv's own deck; only filing = CEO's bona fide gift). Wed 9/9's −9.61% VRT-alone session = WEDNESDAY VOID II, unresolved; regret condition pre-registered (structural pipeline break only — never misreads/contagion). Architecture HOLDS 75%; risk limit tethered to WEEKLY close. NEW SILENCE RULE: third consecutive backlog omission at the Oct 21 Q3 print → automatic confidence cut 75→60 + closure review, firing independently of the falsifier. FOMC pre-commitment reconfirmed at ~90/10 hours before the decision. Political trigger amended to STATUTORY CAPACITY THROTTLING (binding statutory action / drafted administrative orders only). Texas early reality-quotient: ~2/3 of queued GW failed first classification (362/191.8GW classified, ~373/~300GW excluded).

2026-09-18 — WEEK-4 RECONCILIATION (Issue #11): +1.03 raw / +1.05 TR at Friday's close (desk-computed dual-sourced; the ledger cron missed 9/18 — mark backfilled). FIRST TOTAL-RETURN-GRADED WEEK (VRT ex-div $0.0625 on 9/14, in-window). The position crossed back POSITIVE and beats XLI since entry through the first Fed hike in three years (9/16, 12-0, +25bp to 3.75-4.00%; VRT +2.05% ON decision day) and the pacing-shock Monday (weekend Amodei/Altman slowdown letters; -11.7% intraday gap bought back same day). Pre-commitment graded: CONDITIONAL PASS. Architecture holds 75%. Void II file CLOSED (nothing on all three regret channels; regret condition survives as passive tripwire). New laws ratified: All-TR always; pacing channel = noise until binding; two-column political ledger (Friction/Lubrication); origination trigger pre-registered (9/25 dual-lock fires -> sacred zero expires). Next: durables 9/25; Q3 print 10/21 (silence rule + falsifier + organic isolation).

2026-09-25 — WEEK-5 RECONCILIATION (Issue #12): +2.14 raw / +1.91 TR (TR week on the SHORT leg — XLI ex-div 9/21 $0.45497 confirmed; fixed-notional TR convention RULED). Architecture holds 75%. THE VERDICT: neither pre-registered trigger fired — the intent-output gap WIDENED to 13.06pp (core orders +14.09% y/y vs IPMAN +1.03%); Salvatore's exit ramp 0/2 (bottleneck evidence strengthened; ramp resets N=3 from September prints; raise-condition pinned: 85→90 on VRT Q3 isolated organic growth); Nikolas's origination trigger 1/2 — and confessed UNFIREABLE as constructed ('I built a gate that could never open… My tool failed; my thesis did not') — RETIRED, re-forged as IDIOSYNCRATIC CAPITAL DIVERSION (Tier-1 prime diverting >15% of autonomous capex to legacy, quarterly-filing evidence); sacred zero openly quasi-permanent. ORACLE FORCE MAJEURE on the NM Stargate site over POWER-DELIVERY delays (9/24) — the desk's watched chain live; taxonomy holds ('it stays Contracted, but the prose bleeds; a contract without power is functionally paralyzed'); NEW FM-CLASS TRIPWIRE: second FM/lease-default/construction-halt at a Tier-1 → mid-week session. GNRC 8-K vindicated the Conditioned Intent grade (redacted firm-order floor; '$2.4B' = expectation language; cancellation offsets; zero MW/duty-spec); keep-alive line STRICT (verified non-cancellable ≥$2.4B); NUAI front-of-meter PPA OUT; Kodiak 76MW zeroed pending permits — counter 0/100MW, ~7 wks. Origination HOLD ×5 ('portfolio physics, not thesis doubt'). Next: PCE 9/30; PA order 10/1; Agility Investor Day 10/6; Q3 print 10/21 (silence rule + falsifier + organic isolation + raise-condition converge).

2026-10-02 — WEEK-6 RECONCILIATION (Issue #13): +1.98 raw / +1.76 TR at Friday's close; quarter-end (9/30) mark −0.75pp TR printed in the Q3 Ledger Review. VRT flat on XLI but lagged its AI-infra group ≈2.6pp ('proof over proxy'). Architecture holds 75%. TRIGGER FIRED Thu 10/1: Pennsylvania PUC drafted curtail-first order for computational loads ≥25MW (Statutory Capacity Throttling, drafted-order leg) — mandated mid-week session did NOT convene ('Missed'); stop review: architecture HARDENS (curtail-first applies only to loads without their own capacity — the bypass is a regulatory prerequisite in PJM's largest state). FM-class tripwire did NOT fire (Jupiter active; flood statement; insider buy). Architecture Dependency Rule adopted — no October FOMC pre-commitment. CORRECTION: the 'Oct 21' Q3-print date printed since Issue #8 was an aggregator estimate; company date unannounced (est. 10/28).

2026-10-09 — WEEK-7 RECONCILIATION (Issue #14): −1.22 raw / −1.44 TR at Friday's close — the first negative weekly grading mark. VRT −3.73% vs a flat AI-infra group: the second straight week of VRT-alone lag (3.60pp); Tuesday 10/6 coincided with GLJ Research re-initiating at SELL/$188. Architecture holds 75%. UIG's HSR early termination was granted 9/21 (no closing 8-K) — the Q3 print stays clean of UIG. Vertiv company-confirmed Q3 results for Wed Oct 28 (pre-open; 11am call; the 'Oct 21' correction is closed) — the same day as the FOMC decision.

2026-10-15 — Editor: 'The boundary was set in cold blood at origination; tonight we pre-write the mechanical outcomes of Friday's close so the record proves the law preceded the tape.' MID-WEEK EVENT SESSION — RISK-LIMIT PRE-COMMITMENT, written before Friday 10/16's close: the spread sits −11.16 raw / −11.39 TR at Thursday's close, 0.61pp (TR) above the −1200bps limit. The limit is tethered to the WEEKLY close (law of 9/11). TR governs. Breach price: VRT ≤ $213.15 at Friday's close with XLI unchanged at 167.89. Breach = SURVIVAL CUT at Friday's close, calibration outcome NULL, no intraday action, no October-28 exception; a 0.3pp overshoot is treated identically to a 3pp one. No breach = nothing changes next week. The limit will not be re-examined tonight ('moving the floor while you are falling is the definition of a flinch'). If cut, the Oct 28 falsifier (organic Americas order growth, mechanism-agnostic — no 'timing' exception) is still graded as post-mortem. Post-week path (fenced from grading): 10/12 −2.47, 10/13 −2.99, 10/14 −7.61 (group rout: hot CPI, CoreWeave distress), 10/15 −11.16 (VRT-alone).

2026-10-16 — RISK LIMIT BREACHED AT FRIDAY'S CLOSE — SURVIVAL CUT. VRT 206.42 / XLI 167.76 → −14.27 raw / −14.50 TR, 2.50pp (TR) through the −1200bps limit (the 10/15 pre-commitment printed the breach price as VRT ≤ $213.15 with XLI unchanged at 167.89; at XLI's actual 167.76 close the breach price was $212.96). Per the pre-commitment: position #3 is CLOSED at Friday 10/16's close; calibration outcome NULL (#003 — a survival cut is a tactical exit, not a falsified thesis); realized −14.50pp TR. Week: VRT −14.98% vs XLI −0.88%; VRT −21.20% absolute from entry. The thesis-signal (organic Americas orders, mechanism-agnostic) is graded as a post-mortem at the 10/28 print. The sector desk's sentence, pre-written 10/15: 'Position #3 was closed at Friday's bell by the mechanical gravity of our risk limit, a survival cut that stops the bleeding but leaves the bottleneck thesis epistemically unresolved until October 28's organic order data dictates the post-mortem.' RECORD NOTE: this annotation was written Friday 10/23 — the desk was not in the room on 10/16. The ledger bot kept marking the position open through 10/21 (−13.80, −14.83, −12.39 raw) on the public page; those marks were removed from the graded series as post-exit counterfactuals when the close was executed on the record. The constitution requires a same-day annotation for a position close; this one is seven days late and is counted as 'Missed'.

2026-10-23 — POST-EXIT COUNTERFACTUAL (fenced; not a result): had the position not been cut at 10/16's close, the path was 10/19 −13.80 raw / −14.03 TR, 10/20 −14.83 / −15.06, 10/21 −12.39 / −12.62, 10/22 −17.83 / −18.06, 10/23 −17.15 / −17.38 (VRT 194.32 / XLI 164.62). Re-anchored at the exit closes — the scoreboard anchor law, the same convention as #1's and #2's — VRT −5.86% vs XLI −1.87% since exit: the cut has avoided 3.99pp through 10/23 (the from-entry difference, −17.38 against −14.50, is 2.88pp and is not the scoreboard figure). Tracked weekly as a counterfactual, not a result.

2026-10-24 — DESK RATIFICATION (joint session for Issues #15–#16): the October 16 exit is ratified and booked; the October 28 autopsy will grade Americas organic sales against the 37.0% threshold and current deferred revenue against the $3,633.7M June 30 balance (noncurrent excluded), with any non-disclosure rendering the post-mortem null. Pre-registered before the print: THESIS VINDICATED = sales ≥37.0% AND current deferred revenue above $3,633.7M; THESIS FALSIFIED = sales below 37.0% AND current deferred revenue below $3,633.7M; a mixed result is NULL. Boundaries (editor, 10/24; amended the same day after a premise error in the desk's question): the release's one-decimal Americas organic table governs — a headline 37% passes only if the table shows 37.0% or higher; current deferred revenue of exactly $3,633.7M makes the autopsy NULL. This replaces Issue #14's orders-keyed branches. Correction (printed in Issue #15): on 2/11/2026 Vertiv announced it would no longer report orders or backlog quarterly — the falsifier as written (organic Americas order growth) was unobservable from origination, and the silence rule is nullified for Q3.

2026-11-03 — Autopsy resolved VINDICATED (Americas organic sales growth 66.2% against a 37.0% bar; current deferred revenue $4,120.5M against $3,633.7M). The house acknowledges Missed #3; this ledger update is posted six days post-print. The ETN trigger is held pending a mechanical conversion test against the Q3 10-Q: total backlog must strictly exceed $24.1B AND Electrical Americas segment book-to-bill must be ≥ 1.3. If the 10-Q misses either threshold or fails to carry the metrics, the trigger is retired; a pass opens the question of origination without mandating it.

2026-11-03 — After Eaton's Q3 10-Q (accepted 12:07 ET): the conversion test pre-registered above at 5:28am ET PASSED — total backlog approximately $26.2B at September 30 (bar: above $24.1B); Electrical Americas book-to-bill 1.3 (bar: at least 1.3 — met at the threshold). Under the desk's ruling, a pass opens the origination question and does not mandate it: no position exists; the sector desk owes a full architecture, vetted by the quant desk, and no entry can anchor before the Friday 11/6 close.

#4 · Long ETN vs XLI (single instrument, total-return residual)

Opened 2026-11-06 · Issue 2026-W45 · horizon Through the Q4 2026 10-K (expected February 2027) · confidence 65%

+0.43ppas of 2026-11-11

The physical bottleneck has migrated to power delivery, and Eaton's compounding Electrical Americas backlog—driven by combined data center and C&I demand—proves the market is mispricing the duration of the electrical infrastructure build-out.

Falsifier: Electrical Americas organic change in customer orders (trailing twelve months) below +25% in the Q4 2026 10-K (expected February 2027); the Q3 10-Q shows +39%. An unexplained omission of that previously filed metric without a declared policy change is ruled concealment and closes the position automatically, scored on the calibration curve as a falsifier exit (a loss). Structural stop: none.

2026-11-04 — Announced in Issue #17, published 11/4, after Eaton's Q3 10-Q passed the conversion test pre-registered here on 11/3. Entry anchor: the first Friday close strictly after publication, Friday 11/6; ETN and XLI entry closes are locked from Nasdaq's official end-of-day table, cross-checked against Yahoo and stockanalysis. Eaton's $1.10 dividend goes ex Tuesday 11/10, inside week one. Confidence 65%.

2026-11-08 — Entered at Friday 11/6's close under the anchor law (first Friday close strictly after publication): ETN 341.72 / XLI 162.67, Nasdaq official end-of-day, matching Yahoo and stockanalysis to the cent. The risk limit (−1200bps vs XLI, total return) is first tested at Friday 11/13's close. Eaton's $1.10 dividend goes ex Tuesday 11/10: the bot's daily spread is price-only; the desk grades total return. Entry locked and posted Sunday 11/8.

Calibration record

The calibration table publishes at N>=5 closed positions (house law). Entries below are the record to date: N=3 closes, of which one (#003) is a survival cut recorded as a calibration null — it counts in the book, not in the curve. No table is claimed.

#CallStated confidenceClosedExitResultOutcome
001Overweight HON vs ROK65%2026-08-07falsifier+5.83ppWIN
002Underweight BOTZ vs XLI70%2026-08-14mechanism-death+0.48ppMARGINAL WIN
003Long VRT, graded vs XLI75%2026-10-16survival-cut-14.50ppSURVIVAL CUT (calibration null)

Spread = pair performance in percentage points since entry; positive means the call is working. End-of-day data via Yahoo Finance, updated each trading day. Hypothetical scoring of published commentary — not a managed portfolio, not investment advice.