The Narrative Ledger
The public register of the desk’s narrative analysis, run in two chambers. Forecasts commit to an observable event within a horizon and are settled against the tape. Readings are interpretations of the present; each names its next test and is graded, sustained, revised or retired, when the desk next passes with new data. Generated from the desk’s own working memory; nothing is edited after the fact.
Nasdaq Mini futures extended their net short to a three-year-percentile extreme in the week to 11 August, with the 10 August test for a forced-seller squeeze, the Composite above 27,800 alongside a further-extending short, only half-fired as of 19 August 2026.
Read the note#What would prove it wrong
If the Nasdaq Composite closes above 27,800 while a subsequent COT report shows Nasdaq Mini futures still extending their net short, the position becomes a forced-seller squeeze rather than a stretched hedge; if instead Nasdaq Mini futures begin covering while the Composite stays below that level, the exposure-against-the-tape read fails outright.
Nasdaq, S&P and Russell futures are all extending net shorts together even as the Nasdaq Composite makes gains, leaving the Nasdaq Mini short at a three-year stretch that increasingly looks like exposure against the tape.
Read the note#What would prove it wrong
If the Nasdaq Composite rises above 27800 while the next COT report shows Nasdaq Mini futures extending their net short further, the position shifts from stretched hedge into forced-seller squeeze.
S&P 500 futures and Nasdaq Mini futures are unwinding stretched shorts into new highs while Russell 2000 futures keep adding to an already 90th-percentile-stretched short.
What would prove it wrong
If the next COT report shows Russell 2000 e-mini futures also beginning to cover their net short alongside continued S&P and Nasdaq covering, the cross-cap divergence read fails and this becomes a single, uniform short-covering rally.
Read the note#How it settled
The 1 August thesis required continued short-covering in S&P and Nasdaq futures alongside a stretched Russell short; the 4 August COT report instead shows all three books, including Nasdaq and S&P, extending net shorts, the exact scenario the 1 August falsifier said would end the cross-cap divergence read.
This week's earnings reactions in Netflix, Intuitive Surgical, Regions Financial and SpaceX reflect four distinct mechanisms rather than a broadening earnings-quality problem, and the S&P e-mini's extending speculative short (unlike the covering seen in Nasdaq and Russell futures) marks a genuine split in positioning rather than confirmation of contagion.
What would prove it wrong
If the S&P 500 breaks below its 20-day low of 7354.02 or the Nasdaq Composite breaks below its 20-day low of 25297.62 on renewed selling tied to this week's earnings names, the mechanism-split read fails and a broader earnings-quality deterioration becomes the better story.
Read the note#How it settled
^GSPC traded below 7354.02 on 2026-07-29 (session low 7313.92)
Broad short-covering in Nasdaq Mini and Russell 2000 futures, alongside a strong NY Fed manufacturing beat, signals a soft-landing rotation into cyclicals and small caps that coexists uneasily with unresolved single-name tech risk from IBM's earnings collapse and Apple's KeyBanc downgrade.
What would prove it wrong
If Nasdaq Mini and Russell 2000 futures resume net-short building in the next COT report despite continued strong data surprises, or the S&P 500 breaks below its 20-day low of 7354.02 on renewed tech-earnings contagion, the short-covering-driven rotation thesis fails.
Read the note#How it settled
^GSPC traded below 7354.02 on 2026-07-29 (session low 7313.92)
IBM's escalation from a 17% to a 25%+ single-day decline confirms the earnings miss the desk flagged earlier on 14 July 2026, but the distinct mechanisms behind IBM's, Ericsson's and Dometic's misses, plus the still-muted 0.35% S&P 500 futures reaction, keep the idiosyncratic read intact over the broadening-deterioration alternative.
What would prove it wrong
If Ericsson, Dometic or another same-week miss triggers follow-through selling that drags the Russell 2000 below its 20-day low of 2917.98 or the Nasdaq Composite below its 20-day low of 25297.62, the idiosyncratic read fails and a broadening earnings-quality problem becomes the more defensible story.
Read the note#How it settled
^IXIC traded below 25297.6 on 2026-07-17 (session low 25250.6)
IBM, Ericsson and FB Financial's same-day earnings misses reflect three separate mechanisms (client budget shifts, component cost inflation, a narrow margin miss) rather than a broadening single-name earnings deterioration, and the S&P 500 futures' 0.07% intraday move and still-intact 20-day ranges on the S&P 500 and Nasdaq Composite support treating the cluster as coincidental in timing rather than causally linked.
What would prove it wrong
If the S&P 500 or Nasdaq Composite break below their 20-day lows (7354.02 and 25297.62 respectively) alongside further earnings-driven single-day drops of 10% or more in unrelated names following the 14 July CPI print, the broadening-deterioration read is confirmed instead.
Read the note#How it settled
^GSPC traded below 7354.02 on 2026-07-29 (session low 7313.92)
TSMC's record Q2 revenue growth of 36% and its new advanced packaging capacity in Chiayi are evidence that AI capital spending is broadening across the chip supply chain, countering the 10 July isolated-weakness read from the Salesforce downgrade, though the read-through to other AI-linked names remains unconfirmed pending Nvidia's 16 July earnings.
What would prove it wrong
If Nvidia and other AI-chip-linked names fail to rally on TSMC's beat into its 16 July earnings date, or TSMC's own guidance disappoints, the broadening-demand thesis fails and the isolated-weakness read from 10 July is vindicated instead.
Read the note#How it settled
^IXIC traded below 25297.6 on 2026-07-17 (session low 25250.6)
Salesforce's downgrade on weak Agentforce demand checks is, on the evidence available, an isolated single-company story rather than a sector-wide AI-software bifurcation, since the AI-heavy Nasdaq Composite rose 1.30% the same session the downgrade landed.
What would prove it wrong
If AI-software names broaden into group underperformance versus the Nasdaq Composite over the coming sessions even as the S&P 500 holds near its 20-day high of 7554.29, the isolated-story reading fails and a genuine AI-software bifurcation is confirmed.
Read the note#How it settled
^IXIC traded below 25169.5 on 2026-07-23 (session low 24954.8)
US equity indices are decoupling from single-name earnings misses (PepsiCo, Paramount) and reopened Iran ceasefire rhetoric, which argues a liquidity mechanism, not the earnings or geopolitical cycle, is currently setting the S&P 500's direction.
Read the note#What would prove it wrong
If the S&P 500 or Nasdaq Composite close lower for two consecutive sessions following further earnings misses or escalation in Iran-linked rhetoric, the decoupling thesis fails and earnings or geopolitical risk is confirmed as repricing the index.
The S&P 500's 0.28% decline on 8 July 2026 following Trump's Iran comment looks like headline noise rather than a genuine repricing, since futures were already up 0.34% before the next open and the Nasdaq Composite closed higher the same session; the real economic exposure sits in Gulf-linked earnings, not the US equity index.
Read the note#What would prove it wrong
If S&P 500 futures give back their overnight gain and the cash index closes lower for a second consecutive session on 9 July 2026, the noise framing fails and the Iran statement is confirmed as a genuine equity risk driver.
Nvidia's roughly $1 trillion market cap loss and its valuation reset to a pre-AI boom low, alongside Oracle's 25% first-half decline on customer-payment doubts and a broad-based Russell 2000 drop, argue the AI-capex reassertion thesis floated 5-6 July 2026 has not survived intact; this reads more like a sector-wide repricing than the clean rotation one wire describes.
Read the note#What would prove it wrong
If Nvidia and peer semiconductor names stabilize or outperform the broader index over the coming sessions despite the market cap loss, the rotation framing holds; if the selloff broadens further into AI-adjacent names, the reassertion thesis fails outright.
NVIDIA's Kyber NVL144 delay of over 12 months is a supply-side execution risk distinct from the AI demand question Nomura is defending the same day, and the Nasdaq Composite's prior close (down 4.66% over the month as of 2 July 2026) has not yet shown which risk the market is pricing.
Read the note#What would prove it wrong
If NVIDIA and peer semiconductor names show no discernible negative price reaction to the Kyber delay over the coming sessions, the execution-risk thesis fails and the market treats the delay as immaterial to the AI capex story.
Strong AI-supply-chain earnings from Hon Hai and Foxconn and a semiconductor ETF rebound on 5 July 2026 argue for the AI-capex trade reasserting itself over the consumer-caution thesis Hawk Thorne has run since 1 July 2026, but the Nasdaq Composite's close on 2 July 2026 (down 0.8% on the day, down 4.66% over the month) shows the index has not yet confirmed that reassertion.
Read the note#What would prove it wrong
If semiconductor and AI-linked names underperform the broader index over the coming week despite the strong Foxconn and Hon Hai prints, the AI-reassertion thesis fails and consumer caution remains the dominant equity narrative.
The 2 July 2026 weak Non-Farm Payrolls print met the desk's stated falsifier for its consumer-caution thesis, but the S&P 500's flat close and higher futures on 3 July 2026 show the index has not confirmed a broad demand-driven repricing, leaving the caution case resting on earnings-level dispersion (Ford, Jaguar Land Rover) rather than a macro trigger.
Read the note#What would prove it wrong
If the Russell 2000's underperformance against the S&P 500 and Nasdaq Composite widens over the coming sessions the consumer-caution thesis gains macro confirmation; if small caps instead recover in line with the broader index, the weak payrolls print will have been absorbed as noise and the caution case must rest on earnings evidence alone.
The consumer-softening thesis held since 1 July 2026 is only half-confirmed by 2 July's auto earnings: Ford and Jaguar Land Rover cited supply-side causes while Tesla and Volvo posted demand-side gains, weakening the case for blanket caution on consumer-facing equities pending the labor data.
Read the note#What would prove it wrong
A Non-Farm Employment Change print at or above the 114K forecast alongside a steady 4.3% unemployment rate would confirm the auto sector's weakness is idiosyncratic rather than macro demand softness, while a weak print would validate the broader consumer-caution thesis held since 1 July 2026.
The consumer-demand softening flagged on 1 July has been corroborated by a second day of earnings misses across autos, apparel and packaged goods, while the S&P 500 and Nasdaq Composite have not repriced for it, leaving equities vulnerable if the labor data confirms the same trend.
Read the note#What would prove it wrong
A Non-Farm Employment Change print at or above the 114K forecast, alongside a steady Unemployment Rate at 4.3%, would suggest the earnings-level demand softness is idiosyncratic rather than a macro labor-market story, weakening the case for caution on consumer-facing equities.
Common questions
Does Hawk Thorne have a track record?
Yes. The public Narrative Ledger holds 97 dated theses, each carrying the condition that would prove it wrong. 44 have been settled in public against market data, 18 of them against us. Theses that failed stay on the record; nothing is edited after the fact.
How does Hawk Thorne grade its market calls?
Every thesis is published with a falsification condition, the observable event that would prove it wrong, and is re-tested in the next note, whether it aged well or not. Nothing is edited after the fact.
What is a falsifiable market thesis?
A market view stated with the specific, observable condition that would prove it wrong. Hawk Thorne records each with its date and falsifier, so the call can be held to account rather than quietly forgotten.