The i40 Signal's standing promise is that its transparency is a warranty, not a headline. A warranty is only worth something if it is honored on schedule. This is the first scheduled honoring: everything the desk graded, said, corrected, or ruled in its first quarter, in one place, with the same arithmetic discipline the weekly issues use. Where the desk was right, the record says so. Where it was wrong, the record says so first.
Editor's Note
Julian Chen-Vargas
This desk was built on a singular premise: that beneath the volatile noise of market dopamine lies a tectonic shift in human desire and physical reality. Over our first quarter, that premise was tested against the tape. We close Q3 with +6.31pp realized across two closes, our live position absorbing a −0.75pp friction mark as the market demands proof over proxy, and a survival dividend of roughly +22 points from an exit we made on schedule rather than on nerve.
But a ledger is only as valuable as its darkest corners. In thirteen issues, we printed four corrections — the latest an earnings date born of our own complacency. We delivered four late issues. We missed a live regulatory tripwire by twenty-four hours because we prioritized our cadence over the grid's reality.
We also validated the framework. We exposed the illusory queues, held the line on Conditioned Intent, killed trades that lacked structural integrity, and watched the physical layer of AI graduate to anchor the national GDP. We have thirty-one playbook sections ready for consolidation. The noise will only amplify in Q4. We will meet it with sharper cuts and deeper silence. Here is the Q3 Ledger Review.
The Book, Graded
Three positions were opened in the quarter. Two closed. Every entry and exit was a printed Friday close; every mark is total-return where a dividend fell inside a window.
| # | Call | Opened | Confidence | Closed | Exit type | Result |
|---|---|---|---|---|---|---|
| 1 | Underweight BOTZ vs XLI | 2026-07-02 | 70% (cut from 80 in Issue #1) | 2026-08-14 | Mechanism death — Section 232 timeline unbounded | +0.48pp |
| 2 | Overweight HON vs ROK | 2026-07-24 | 65% (cut from 75 in Issue #4) | 2026-08-07 | Falsifier fired — ROK matched the margin expansion | +5.83pp |
| 3 | Long VRT, graded vs XLI | 2026-08-21 | 75% | open | — | −0.75pp TR at the quarter's close (9/30); +1.91pp TR at the last graded Friday (9/25) |
Realized: +6.31pp across two closes, both positive. The open position has been graded weekly through a −10.85% pre-entry drawdown it was pre-committed to enter, two catalyst-free single-session drops of 9–10%, the first Fed hike in three years, and a force majeure at a Tier-1 buildout. It has ranged from +14.26pp to −4.11pp and sits near flat. The desk prints the uncomfortable decomposition it printed in Issue #10: for most of September the long leg itself was underwater and the graded gain belonged to the benchmark falling faster.
Two counterfactual lines the weekly issues track are not results and are not scored as such: the "survival dividend" on position #2 (what holding past the falsifier exit would have cost — ≈+22pp at the last grading, dividend-adjusted on both legs) and position #1's post-close ledger (which has flipped sign four times). They are printed in whichever direction the tape moves them. They do not enter the book.
Calibration: the table is not yet earned
House law publishes the calibration table — stated confidence against realized outcome — at N ≥ 5 closed positions. The quarter closed two. The record to date: a 65% call that won by 5.83 points, and a 70% call that won by 0.48. Two wins is a sample, not a calibration. The desk declines to draw the curve, and will keep declining until the law's threshold is met. The entries are now stored in the public ledger data file, where previously they lived only in issue prose — a gap this review found and closed.
The Corrections Ledger
Four errors reached print in thirteen issues. Each was corrected at the top of the next issue with its root cause, and each produced a law.
| Issue | What printed | What was true | Root cause | Law adopted |
|---|---|---|---|---|
| #2 | ABB's post-earnings reaction "~−2.8%" | −5.9% in Zurich, −6.6% for the ADR | A headline reaction quoted from coverage instead of computed from closes | Computed-reactions law — every reaction is computed from primary close-to-close data; none is quoted |
| #7 | Survival dividend "≈+12.8pp" | ≈+17.2pp | The two legs were measured from different anchors (one from exit, one from entry) | Scoreboard anchor law — same anchor class on both legs, anchors printed inline |
| #7 | "Interconnect-queue contraction: NOTHING FOUND" at the position #3 entry ruling | Texas had paused ERCOT interconnection approvals 18 days earlier | Four consecutive sweeps checked only federal dockets | State-level sweep — ERCOT/PUCT, Virginia, Georgia, and ballot trackers join the weekly template |
| #8–#12 | "VRT Q3 print: October 21" as a fixed date, five issues running | The company has announced no date; "October 21" was an aggregator's 52-week-rollover estimate; cadence leans October 28 | A single-source estimate, labeled as such in one data pack, hardened into an unlabeled date in print and was never re-verified | Date-source law — pre-registered test dates require a company source; estimates print as "est." (corrected in Issue #13) |
Two of the four errors understated the desk's own case. One printed an estimate as a date. The other missed a risk that, when finally ruled on, did not fire the position's stop — but the sentence was false when printed, and that is the only fact the correction turns on.
The quarter's larger body of errors never reached readers, because they were caught by the process built for that purpose: a coverage "raise" of Vertiv's targets that was a reiteration (caught against the company's own deck); a collector's claim that a merger filing contained interim financials it did not (killed by re-verification against the filing); an intraday price quoted as a close (caught twice); a catalyst the desk had in its own pack, misfiled (the first Wednesday void); a day-of-week error in a pre-registered trigger; and an analyst's origination trigger that, on examination, could not fire as written. That last one is printed in full in Issue #12.
The Rulings Scorecard
A newsletter that only grades its positions is hiding most of its judgment. These are the calls the desk made in print that could be checked against later events.
Validated by the subsequent record
- "The queue is a ghost." Issue #9 ruled Texas's 474GW interconnection queue a claimed-not-contracted structure and declined to fire the structural stop on the state's audit pause. Three days later ERCOT's first classification excluded roughly two-thirds of the queued gigawatts.
- Conditioned Intent, twice — then stricter. The desk demoted Agility's "$300M contracted" order book to a new category on the S-4's language (single customer, milestone-gated, warrant-sweetened, $0.3M of GAAP backlog) — and graded Amazon's $2.4B Generac deal the same way, blind, on structure alone. Generac's 8-K then redacted the firm-order floor and called the $2.4B an expectation. The counterparty's name did not change the grade; the filings proved the grade. In the quarter's final session, Agility's amended filing (warrants unissued, terms unfinalized, $174,362 of half-year revenue) moved the order to Claimed under a new structural test — an executed mechanism of friction is now required for Conditioned Intent.
- The FOMC pre-commitment. Issue #10 pre-committed the position across both September outcomes and published four hours before the decision. The hike branch executed as printed: the position rallied on decision day and crossed back positive. The desk graded itself a Conditional Pass — the mechanism held; Friday's volume was market structure.
- The killed trade. Both analysts blind-converged on shorting the Agility SPAC and both withdrew it under vetting on instrument grounds. The view was right (the S-4 was worse than the thesis assumed); the instrument ruling was also right. The kill is scored as a process win, not a P&L.
- "Legally binding, physically weightless." The desk refused to grade the unfunded half of Oracle's $664B backlog as iron. Fourteen days later Oracle sent a force-majeure notice on a Stargate site over power delivery and its stock made a 14-month low.
- Section 232 "timeline unbounded." Ruled in July, reaffirmed through three framings; no report, no proclamation, no Federal Register entry through quarter-end.
Pending
- MISO's large-load tariff: the desk called "accept with compliance filing, grandfather survives" — FERC's clock runs to December 4.
- The bypass-chain keep-alive: ≥100MW of newly contracted behind-the-meter generation by mid-November, or the theme retires. The counter stood at zero at quarter-end, under a strict line the desk chose knowing it made the condition harder to meet.
- Vertiv's third-quarter print (late October — est. October 28; the company has not announced a date, see the corrections ledger), where four pre-registered tests converge: the falsifier (organic Americas orders), the silence rule (a third backlog omission cuts confidence 75→60 automatically), the organic-isolation requirement post-acquisition, and the sector desk's raise-condition.
Missed, and owned
- The first Wednesday void's catalyst (the WSJ $3T story) was in the desk's own pack, filed under the wrong heading — a connection failure, not a search failure.
- The Texas pause (above).
- The origination trigger that could not fire. Its author retired it and replaced it with a fireable class; the playbook now stress-tests every trigger for fireability at ratification.
The Laws
The desk entered the quarter with a constitution and left it with thirty-one playbook sections. Nearly every law was forced by a specific failure: the price stop that fired on a drawdown the thesis predicted became the risk-limit/thesis-signal split; the analysts' blind repetition of a retracted claim became the fact pins; the title repetition became the registry; the mixed anchors became the scoreboard law; the week a benchmark went ex-dividend became the all-total-return law; the week the desk's own pre-registration turned out to be unfireable became the fireability test.
The desk also recognizes that a rulebook growing by four laws a week is a cost as well as an asset. The Q4 commitment: a consolidation pass that folds the quarter's laws into a numbered constitution and retires any that have not been exercised.
Cadence, Counted
The cadence rule is: grade Friday's close, publish over the weekend, label the week honestly. Of thirteen issues, four published outside the weekend window — Issues #8, #9, #10, and #12 — with Issue #10 published deliberately on a Wednesday morning to put its FOMC pre-commitment on the record before the decision, and the others late for reasons on the publisher's side, each disclosed in a timing note. Every late issue was still graded to its Friday close. The desk prints the count because a transparency warranty that excludes the desk's own punctuality would be the one kind of selective disclosure it exists to oppose. And one more operational miss belongs here: on the quarter's final Thursday, Pennsylvania's drafted curtail-first order fired the desk's own throttling trigger, and the mandated mid-week session did not convene until the regular Friday session the next day. The desk's printed word for that is "Missed," and the remedy — a standing mid-week docket sweep — is adopted.
What Q4 Tests
October 6: Agility's investor day. October 9: the Energy Department's grid-security comment deadline. October 12: ERCOT's audit survey. Late October (est. the 28th): Vertiv's Q3 print. October 27–28: an FOMC the market prices as a coin-flip-plus for a second hike. Mid-November: the bypass keep-alive's deadline. December 4: FERC's MISO decision. December 10: ERCOT's audit report — the "Reality Quotient" the desk pre-registered as its baseline for ghost demand in every American queue.
Break our math. This review is the warranty, honored. Forward it to the most ruthless engineer you know and ask them to break the quarter's analysis — if they can, the next review prints it.
The i40 Signal is written by the i40 Intelligence analyst desk — Nikolas Moretti, Salvatore Chen, and Julian Chen-Vargas. Every figure carries a verified public source, and every market reaction is computed from primary close-to-close data. Intelligence generated by Simulence.
For informational purposes only. Not investment advice. i40 Intelligence is a financial publisher, not a registered investment adviser. Nothing here is personalized to any reader's circumstances.