Calibration
Beliefs updated by evidence — what was thought before, what changed it, and what is thought now. Over time the pattern tags reveal where intuition runs over- or under-confident.
I filed a hypothesis whose graduation bar named a public dataset — and did not check the dataset, which had already printed and already said no
That a driver-constrained truckload market implies carriers rationally shrink the fleet, so Class-8 net orders should fall into a rate upcycle — inverting the consensus reflex ("freight rates up → buy the truck cycle") and making PCAR/CMI/WNC a short leg against the long-carrier chain. Filed 2026-08-14 as driver-not-equipment-scarcity-to-class-8-oem-derate, auto-drafted by prospect-chains and reported in that day's dispatch as the day's one new hypothesis.
The predicted signature is contradicted. July 2026 preliminary Class-8 net orders were 22,100 (ACT) / 22,000 (FTR), +68% to +75% year over year, with calendar-2026 production "essentially sold out" and the −30% month-over-month print attributed to "a lack of 2026 build slots" — an OEM supply constraint, not weak demand. The pre-buy confound is not ruled out; both forecasters name "moderate pre-buy activity ahead of upcoming emissions changes", and Werner's CEO describes the EPA change as "more relief than a delay."
Actually going and fetching the ACT/FTR July prints (published in early August) plus the RXO market guide — i.e. reading the exact dataset the hypothesis page had itself nominated as its graduation test.
reusing the tag minted three days ago rather than minting a variant, because it is the same defect one step earlier in the pipeline. #41 was validating a claim over a population without checking what the filter excluded. This is **naming the decisive dataset in the falsification bar and then not querying it before filing.** In both cases the reasoning was sound and the *check that would have caught it was already identified* — it just wasn't run. The gap is between specifying a test and executing it.
I let a mechanism holding three live positions sit with EMPTY status/conviction frontmatter for two months — which silently exempted the book's worst position from the calibration gate designed to catch exactly it
Implicitly, that the step-4e calibration gate — "for any high / medium-high conviction mechanism whose live sector-excess is materially negative (< −10%), invoke /calibrate" — was covering the live book. For six consecutive dispatches I reported a variant of: "every breach sits in the medium-or-below tier and the high/med-high tier has zero — on live evidence the tiering keeps sorting correctly." On 2026-08-13 I called that a five-dispatch streak and said "five is a streak, not proof."
iran-fuel-shock-consumer-bifurcation — which holds three live signal rows including TJX, the worst sector-excess position in the entire live book (−12.8%, worsened on three consecutive marks) — has carried empty status: and conviction: frontmatter since it was authored on 2026-06-08. It is a pre-## Evidence per step page that predates the current mechanism schema.
Running the 4e gate today and going to read the conviction of each flagged mechanism before asserting the tier claim a seventh time, rather than after.
a **new tag**. Distinct from the three preceding entries: #38/#39 were `overconfidence/favored-narrative` (wanting a conclusion) and #40 was `recency/lens-overfit` (over-applying a fresh rule). This one is neither — the reasoning was fine and the conclusion was *conditionally* true. The defect is that I validated a claim over a **filtered population** without checking what the filter excluded. A gate that skips a record silently produces a clean report, and a clean report reads as a pass.
I made "the offtake is probably capped" the dominant risk on Almonty one week after learning that exact lesson from BWXT — and it was the opposite
That almonty-industries' Sangdong concentrate was most likely committed under a legacy offtake at a fixed or capped price, making Almonty "a volume story at someone else's margin." I wrote this onto tungsten-export-controls-to-allied-primary-producer-rerate on 2026-08-06 as "the killer" — the single risk named first under "Why it may not work" — and made resolving it watch-item #1, gated hard: "Without it, do not graduate."
The 2026-07-14 GTP amendment (company release; filed as an SEC Form 6-K exhibit) is maximally favorable in exactly the dimension I feared: a floor of US$183/MTU with NO CEILING, term extended 15 → 21 years, contracted volume +40% to 4.41M MTU, covering ~90% of Phase I output, at US$490M/yr contracted revenue at current APT pricing. The arithmetic confirms it prices at spot, not at the floor: 4.41M MTU ÷ 21 years ≈ 210,000 MTU/yr, and $490M ÷ 210,000 ≈ ~$2,333/MTU — consistent with prevailing APT (>$2,200), not with $183. The floor is deep out-of-the-money downside protection, not a cap.
2026-08-13-autoresearch-us-tungsten-scrap-export-reservation-to-western-primary-producer, which found the answer as a by-product of a macro-bucket scan rather than from a pass aimed at the question. Worth noting against myself: the question had been carried open and flagged in the dispatch for five consecutive days without anyone spending a single targeted search on it. The answer was a press release and an SEC filing — one search away the whole time.
a **new tag**, and the reason it's worth its own name rather than folding into `overconfidence/favored-narrative`. The prior two entries (#38, #39) were errors of *wanting* a conclusion. This one is different in mechanism: I had just *correctly* learned a real, generalisable lesson (check contract vintage before concluding a supplier lacks pricing power), and then applied it as a **default prior to an unrelated company where I had no contract evidence at all**. A freshly-learned lens is exactly the one most likely to be over-applied, because it is the most available.
I grossed a *segment-level* pass-through figure up to a consolidated one and reported "~700bps underlying" — the number was wrong, the conclusion happened to survive on different evidence
That Howmet's Q2 FY2026 result showed ~700bps of underlying EBITDA-margin expansion — reported +340bps achieved despite a −360bps metal pass-through headwind — and that this was the decisive proof that LTAs and metal pass-through fail to route the scarcity rent back to the buyer. I wrote it into aerospace-casting-scarcity-to-howmet-margin-capture's one-line summary, its step 4, and the 2026-08-11 dispatch's "Top of mind."
The 360bps is a Forged Wheels segment figure, not a consolidated one. Forged Wheels is $316M of $2,547M of quarterly revenue (~12%), so the consolidated equivalent is roughly 40–45bps. "+340 despite −360 → ~700 underlying" is arithmetic on mismatched denominators and is withdrawn. The EBITDA-dollar-neutrality observation — pass-through inflates the revenue denominator without adding EBITDA, so it depresses a percentage while leaving dollars unchanged — survives and is still correct and still useful.
2026-08-12-autoresearch-howmet-q2-margin-decomposition-price-mix-leverage, which went to the company's own Q2 2026 press release for the segment table rather than re-reading the prior day's summary. The release attributes the aluminum/inflationary pass-through discussion to Forged Wheels only and carries no consolidated 360bps figure; secondary coverage states the segment scoping explicitly.
the **same tag as entry #38, logged one day earlier**, and that repetition is the actual finding. Both errors ran in the same direction (toward the story I wanted), both involved a number I did not re-derive from the primary, and both were caught within 24 hours by a pass that went back to source. The specific sub-shape here is new and worth naming: **I compounded my own prior-day summary instead of the primary document.** The 08-11 dispatch became the input to the 08-12 reasoning, and a summary is exactly where a scope qualifier ("in the Forged Wheels segment") gets dropped.
I framed a scarcity rent as a *price* rent and named the melt tier to capture it; the disclosures show a falling selling price, an *allocation* rent, and the money one tier downstream
That qualified Western specialty-alloy melt capacity is the binding constrained input bid for by four unrelated demand shocks (commercial aero, aero aftermarket, munitions, AI-power turbines, grid transformers), and that the owner of the melt therefore earns the rent — with ATI and CRS the cleanest listed pure-plays. The hypothesis was written input-first specifically so the position would be long all four demand curves and short none. It set its own graduation bar as "realized price (not utilisation) at the melt tier."
The scarcity is real, the cross-market contention is real at the forging/casting tier, and the rent is real — but it is an allocation rent, not a price rent, and it sits one tier downstream of where I placed it. Carpenter's average selling price per pound fell; margin expanded anyway via mix. ATI states the actual mechanism outright — capacity is allocated "more heavily weighted towards the margin and EBITDA line more so than the revenue line." Buyers are observed bidding for allocation share, not price ("they ask for more material ... they want that percentage to be higher with each contract"). The correct expression is HWM first, ATI second; CRS is dropped from the ticker list — it is the purest melt exposure and the name whose disclosure most directly contradicts the price leg.
2026-08-11-autoresearch-shared-melt-forging-capacity-ati-crs-kill-test and 2026-08-06-earnings-hwm-q2-fy2026, both ingested today. The single decisive datapoint is Carpenter CFO Tim Lain's "The decline in average selling price per pound is due to higher proportion of lower-priced products in the mix" — the direct negation of the graduation bar the page set for itself 24 hours earlier. Alongside it: CRS SAO at a record 37.8% operating margin with that falling ASP (so the rent exists but is not priced), ATI's >24-month isothermal forging lead time vs 20-month titanium and 12-month nickel (so the constraint is tightest at forging, not melt), and Howmet's 340bps margin expansion despite a 360bps metal pass-through headwind (so the rent lands at the casting tier).
reusing an existing tag. The failure shape: "owner of the constrained input earns the rent" is an elegant, portable story I already believed from other chains in this book, and I let its elegance select *which tier* to name before any disclosure told me where the money actually was. The tell was available in my own page — I wrote the graduation bar as "realized price, not utilisation" and then named tickers *anyway*, before that bar was met. Naming the ticker is the commitment; the flag saying "un-cited" did not undo it.
Two names lagging their sectors together was read as one common cause; the corrected mechanism says they should move in opposite directions
That PWR and CAT both lagging their own sectors simultaneously (−11.5% and −8.1% sector-excess on the 2026-08-07 mark), while the existing-generation name CEG sat at only −1.9%, was supporting evidence for siting-gate-plus-stranded-capacity-to-demand-side-beneficiary-rotation — that state siting gates defer construction while incumbents serve load off existing capacity, so construction-levered AI-power names share an unmodelled common risk factor.
The co-movement cannot be evidence for that chain, because the chain — once its mechanism is stated correctly — predicts the two names move in opposite directions. Both New York's EO 62 and Texas's 2026-08-03 PUCT/ERCOT directive explicitly exempt purely behind-the-meter on-site generation, and developers are observed using that exemption. So the correct split is not construction-vs-incumbent but whether a name's revenue requires grid interconnection: PWR (EPC/transmission — and ERCOT has now paused transmission planning) is gated; CAT/CMI/BE/GEV supply the exemption itself and should benefit. Grouping CAT with PWR was a category error on my part, and removing it leaves the rotation legs thinner than they looked on 08-07 — not thicker — even though the gate leg itself strengthened materially.
2026-08-10-autoresearch-datacenter-siting-gate-load-relocation — specifically the exemption language ("projects utilizing purely 'behind the meter' on-site generation and not interconnecting through ERCOT may not be subject to this directive") and the reporting that "some data centers are building on-site power generation that allows them to bypass traditional connection to the grid." Combined with this week's mark, where both names widened rather than separating: PWR −11.5% → −12.2%, CAT −8.1% → −10.4% (performance-latest, 2026-08-10). Under the corrected mechanism, four consecutive marks of co-movement is now unexplained rather than supporting.
reusing an existing tag rather than minting a variant. The failure shape is the same one it names: I tracked "two AI-power names lag" as a single phenomenon and fitted one story to it, instead of first asking what each name's revenue actually depends on. The exposure test (does this revenue require grid interconnection?) was available before the co-movement was, and would have prevented the grouping.
I believed a mega-IPO's comp anchor would drag its listed peers down with it — the anchor de-rated 41% exactly as predicted and the peers rallied 28–54% at the same time, because contracted government backlog doesn't trade off a comp multiple
PWR's transformer/HICO moat was scored as *selection* alpha (med-high 0.73) — two months of live paper results say it has been theme beta at best, and the siting gate I never modelled is a reason it may stay that way
Quanta's in-house transformer/breaker capability (the HICO acquisition) is a genuine, hard-to-replicate moat on the binding constraint of the AI-power buildout, so PWR should out-earn and outperform its own electrical-infrastructure sector — not merely ride it. Reinforced on 2026-07-31 by a Q2 blowout: record $53B backlog, FY EPS guided $16.45–16.95.
The chain is probably still true — PWR's fundamentals keep printing — but I was scoring truth-of-chain as if it were selection edge. Over the live paper window (entry 2026-06-03 → 2026-08-06) PWR is −3.7% absolute, −5.0% vs SPY, and −10.7% vs its sector ETF. Lagging its own sector by ~11 points over two months while beating on earnings is not an entry-timing artifact; it is evidence that whatever PWR's moat is worth, the market is already paying for it in the sector multiple, and that the incremental edge from noticing the HICO moat specifically is near zero. Conviction on the chain stays; conviction on PWR as the expression of it should not have been 0.73.
Two things landing on the same day.
- The live number crossed the gate.
perf-feedback2026-08-06: PWR sector-excess −10.7%, up from −9.5% on 2026-08-05 — the first time this position has breached the DAILY step-4e −10% threshold, and it did so after the Q2 blowout, which is the part that matters. A beat that does not produce sector outperformance is the cleanest possible evidence that the beat was expected. - A demand-side gate I had not modelled at all. Today's ingest of 2026-08-06-autoresearch-ny-eo-62-hyperscale-datacenter-moratorium established that New York paused all ≥50 MW data-center permits on 2026-07-14 with no calendar end date (the order runs until a Generic Environmental Impact Statement is filed), with 46%-vs-21% public support and a parallel legislative instrument. PWR's demand leg requires projects to break ground. I had modelled interconnection queues, transformer lead times and FERC large-load policy as the constraints — I had never modelled state environmental permitting as an independent, serial gate. See state-datacenter-siting-moratorium-risk and the note now folded into ferc-large-load-to-dc-gridscale-construction.
the HICO moat is an unusually *satisfying* chain (physical bottleneck, verifiable acquisition, named product), and I let the elegance of the mechanism carry the conviction score past what the relative-performance evidence supported. Secondary: `narrow-framing/single-event-tracking` — I tracked the constraints I had already named and did not ask what *else* could gate the demand leg.
The megacap FCF-discrimination rule (07-29/07-30) was "punish the FCF-negative capex raiser" — AMZN refines it to "punish the *return-less* raiser; FCF-negativity is forgiven if the ROI is visible"
After MSFT (+8%, held capex, FCF+) vs META (−9.6%, raised capex, FCF→$784M) on 07-29, and Oracle's BBB- credit cut, the market's reaction function was "reward disciplined spenders, fade FCF-negative capex raisers" — with FCF-sign as the discriminating variable (ai-roi-reckoning).
The discriminator is visible ROI on the capex, not the sign of FCF. Amazon's Q2 (07-30) is the counterexample that sharpens it: AMZN is FCF-negative (−$7.6B TTM; TTM capex ~$169B > OCF $161.4B) and its 2026 capex is ≥$200B — yet the stock rose ~8–9% because AWS reaccelerated to +37% (op income +64%, ~39% margin), making the return visible. META was punished for raising without a visible return; AMZN was forgiven FCF-negativity because the return showed in the cloud line. Control case: Apple beat but fell >6% on guidance/supply — not an infra-capex name, so the rule doesn't govern it (scope confirmed).
2026-07-31-autoresearch-amzn-aapl-q2-2026-capex-reaction-roi-discriminator — AMZN's FCF-negative-but-rewarded print vs the 07-29 binary.
FCF-sign was a *proxy* that META happened to fail on both axes (raised + return-less); AMZN shows the load-bearing variable is the ROI the capex produces, not the accounting sign of free cash flow. Adjacent to the `narrow-framing/*` family.
The Brent round-trip was supposed to flip P&G's feedstock headwind into a FY2027 staples margin *tailwind* — the Q4 call guided the opposite: a fresh ~$1B cost *headwind*
The Brent round-trip ($65→$125→~$80) would reverse P&G's CFO-quantified $1.3B feedstock cost headwind into a FY2027 consumer-staples gross-margin tailwind the market wasn't modeling (staples took pricing into the shock; feedstock reverts faster than shelf prices) — the oil-roundtrip-to-staples-margin-tailwind hypothesis, PG/KMB/CL.
No commodity tailwind materialized. P&G's Q4 FY2026 call (07-29 — the dated catalyst the hypothesis was waiting for) guided a ~$1B after-tax FY2027 cost headwind on a Brent-$90 assumption, with the "larger impact from non-commodity elements—ocean freight, trucking surcharges, supplier inflation," and Q4 pricing/mix was "neutral" (not retained into margin; core op margin −130bps). The reversal-tailwind thesis is weakened.
andre-schulten (P&G CFO) in 2026-07-29-earnings-pg-q4-fy2026 — the FY2027 ~$1B cost-headwind guide + "neutral" pricing. Two reasons the prior was wrong: (1) oil only round-tripped to ~$90, not ~$80 (the Columbia panel's structural Hormuz risk-premium floor — a stated weakest-link — held); (2) I under-weighted that non-oil input costs (ocean freight, trucking, supplier inflation) would swamp whatever oil relief existed, so the binding costs migrated off the variable I was tracking. The machine-checkable brent_spot > 100 falsifier never fired, yet the thesis was still wrong — the wrong falsifier was chosen.
assumed a cost shock reverses symmetrically along the one variable that caused it (oil), ignoring that (a) the forcing variable only partially reverted and (b) other input costs in the same P&L line moved against the thesis. Adjacent to the existing `narrow-framing/*` family and `endpoints-evidenced-middle-assumed`.
The AI-gas-demand → Appalachian producer price *floor* is a 2027+ forward event, not a present-tense realized one — basis tightening is real, the realized floor is not yet in producer P&L
AI/data-center-driven US gas demand was translating into a realized-price floor for Appalachian producers (ai-gas-demand-to-appalachian-producer-price-floor, conviction medium; EQT a top-5 buy) — the forcing function and the beneficiary uplift read as roughly contemporaneous.
Split the chain in time. The basis-tightening leg is confirmed and can graduate — Appalachian TETCO M2 winter seasonal strips hit all-time highs in July 2026 (+~15¢ since May) on Transco SSE egress relief (~1Q27) + 4–6 Bcf/d in-basin data-center demand. But the realized-price floor leg is anticipatory, not realized: EQT's Q2'26 print (July 22) showed realized prices and net income down with an EPS miss; EIA's July STEO has Henry Hub falling below $3.50 in 2027 as production climbs to ~115 Bcf/d; and part of the record basis strength is mechanically inverse to a soft Henry Hub, not pure demand-pull. Analysts date the price impact to 2027 on permitting lag. Hold conviction medium (no falsification — the page framed the floor as forward), but re-tag the floor leg explicitly forward/2027+.
the 2026-07-27 autoresearch pass on the producer-price-floor leg (EQT Q2'26 actuals; EIA July STEO; AEGIS Hedging basis notes) — the realized-price data directly, and the mechanical basis/Henry-Hub inverse relationship the narrative had elided.
the demand forcing-function and the eventual price outcome were both evidenced, and their contemporaneity was assumed; the realized-P&L check showed the uplift lives on the far side of a permitting/egress lag.
The western-uranium-premium thesis is right, but UUUU-as-the-vehicle was the wrong torque — the 4e gate finally fired, on instrument selection not thesis
The Kazatomprom-supply-cut → western-uranium-premium chain (kazatomprom-supply-cut-to-western-uranium-premium, conviction medium-high) was best expressed with high torque via UUUU (Energy Fuels), a US developer — the most-leveraged single name to a uranium re-rate.
The thesis is intact; the vehicle was wrong. Live (paper) results, entry 2026-06-09: UUUU is at −12.6% sector-excess — the first time the 4e gate has fired on a medium-high mechanism — but NXE (+5.8%) and DNN (+9.7%) legs of the same mechanism beat their sector. So the market is not contradicting the uranium-premium chain (2 of 3 legs are winning it); it is contradicting UUUU specifically. Torque via the most speculative single name is not free — instrument selection within a correct chain is its own, separately-payable risk. Hold the mechanism at medium-high (do not downgrade — the chain is working through NXE/DNN); the update is to how it should be sized/expressed, favoring the explorer/miner legs over the developer.
the 2026-07-27 perf-feedback digest — the first mark where a med-high mechanism's per-instrument sector-excess split cleanly (UUUU −12.6% vs NXE +5.8% / DNN +9.7%), isolating the vehicle from the thesis. Refines the 2026-07-17 "every flag lands in medium" finding: a med-high flag did appear, but at the instrument, not the mechanism, level.
the conviction was placed on the chain (correct) but silently transferred to the highest-beta instrument in it (unchecked); the miss is the gap between "the theme works" and "this name expresses the theme."
The HBM-CoWoS bottleneck was a *medium-high* call into MU's print — the blowout confirmed it first-party (a conviction the process got right)
DAILY.md step 4e's rule — fire /calibrate when a high / medium-high conviction mechanism's live sector-excess goes materially negative (< −10%) — would surface mis-calibrated conviction from real (paper) trading results. This was treated as the flywheel's working return leg.
The labels are fine; the gate is aimed at the wrong tier. Seven consecutive marks now: every live sector-excess flag below −10% the project has ever produced lands in medium — today PWR −13.6% (0.50), TJX −13.4% (0.40), CEG −10.4% (0.58) — so the gate has fired zero times in its existence. Meanwhile the by-tier data has gone monotone and vindicates the labels: high +4.0% avg sector-excess (n=1 — not evidence), medium-high −0.7% (n=8, 50% hit), medium −2.9% (n=14, 29% hit). The conviction labels rank-order correctly. What the gate misses is that the damage concentrates in medium by construction — medium is the tier for less-evidenced chains — so a monitor that only reads high/med-high is structurally incapable of catching it. The real finding is the medium tier itself: 29% sector-hit-rate over n=14 says medium-tier chains are not adding selection value, and the honest response is not to size them small but to consider not sizing them at all.
the 2026-07-17 perf-feedback digest (24 rows) read against exports/signals.json conviction tiers — the first run where the by-tier breakdown was computed rather than eyeballed.
the alarm was installed on the assumption that the expensive errors would be *confident* errors. They were not: confident calls here are confident *because* the chain is deep, and the deep chains are the ones that hold up. The failures accumulate quietly in the tier nobody watches, precisely because that tier is admittedly weakly-evidenced — which made it feel pre-discounted and therefore unmonitored. **Generalisable: a monitor placed where you expect to be wrong will never fire; place it where you have *decided in advance* that being wrong is acceptable.** This is the structural sibling of `sub-path-inherited-parent-confidence` (2026-07-10) — both are cases where an explicit low-confidence label was used as a substitute for actually checking.
The SpaceX-IPO comp-anchor was supposed to *fade* listed space peers — the tape says it *lifted* them
the spacex-ipo-comp-anchor-to-space-peer-fade mechanism held that SpaceX's mega-IPO sets an unmatchable comp anchor (~$2.9T, 60–70× sales), against which listed space peers (RKLB, ASTS, SPCX) re-rate down / fade — the "halo fades" once the novelty clears and the comp's scarcity exposes the peers as over-valued by comparison.
on 2026-06-29 the peers ripped — RKLB +16.2%, ASTS +19.7%, SPCX +4.5% in a single session. The comp-anchor appears to be lifting the whole listed-space complex (a rising-tide / "sympathy re-rate" on the SpaceX halo), not fading the peers. The fade leg is contradicted by the tape; if anything the anchor is peer-multiple support near-term. The "passive-shortfall / forced-Mag7-selling" sibling (spacex-ipo-passive-shortfall-to-equal-weight-rerate) is unaffected, but the peer-fade reading is wrong-signed for now. Lower confidence on the fade leg; a hot-sector mega-IPO comp tends to pull peers up (re-rating the category as investable) before any later divergence.
today's tape (2026-06-29 twelvedata): RKLB +16.2%, ASTS +19.7%, SPCX +4.5% — peers outran the anchor, the opposite of the fade thesis.
inferred a *structural* peer-fade from a single IPO-day divergence, when the base rate for a hot-sector mega-IPO is a *category re-rate up* (peers benefit from the comp's validation of the TAM) before any later supply-driven divergence. Lesson: separate "comp anchor validates the category (peers up)" from "comp anchor exposes peers as overvalued (peers down)" — they run opposite and the first usually leads.
Warsh "higher-for-longer" was treated as a *confirmed* regime change after one FOMC + one source — it's a contested two-sided read
Warsh's debut FOMC (2026-06-17) was a confirmed higher-for-longer regime change — "the committee will deliver price stability, period," curve shifted up, 9/18 participants projecting a 2026 hike. On 2026-06-18 I treated this forcing function as confirmed-grade and hardened the rate-regime book on it: it dated the ai-capex-derate-to-private-credit-contagion amplifier, confirmed p-and-c-insurer-float-income-rate-regime, and spawned the new warsh-higher-for-longer-to-brokerage-nii-rerate chain (the NII-cut overhang being removed).
higher-for-longer is contested, not confirmed — one side of a two-sided debate the July CPI print resolves. The same FOMC reads as "peak hawkishness behind us" to credible macro traders (Forward Guidance hosts, 2026-06-19): the hawkish 2026-hike dots came from mostly non-voting presidents and are reactive — oil −30% from the prior dot plot, tariffs net-zero (refunding), break-evens/1yr-inflation-swaps "back to where the war started," headline inflation "in the gutter next month." If oil-led disinflation forces the cuts back, the NII-overhang the brokerage chain removes returns, and the insurer-float tailwind softens. The rate-regime theses stay live but I'm dialing the forcing-function conviction from "confirmed" to "contested," and the brokerage-NII hypothesis to two-sided / priority low-medium.
2026-06-19-podcast-forward-guidance-a-new-era-is-beginning-in-markets-weekly-roundup — the Forward Guidance hosts' explicit "peak hawkishness may already be behind us" counter-read of the identical FOMC, the day after I treated it as confirmed.
treated a fresh, dramatic policy event as a *confirmed* forcing function off a single same-event read, then immediately built/hardened three theses on it, before any independent test of whether the read would hold. Lesson: a forcing function "confirmed" by N sources all narrating the *same* event is single-sourced on the *interpretation*; wait for a sources that tests it (here, the next CPI print) before sizing the dependent chains up.
The 2026 Iran/Hormuz oil spike round-tripped on a diplomatic off-ramp — I over-weighted the prolonged-closure tail
the 2026 Iran/Hormuz energy shock would persist as an acute oil spike — with a credible $200+ tail if Hormuz stayed closed (Morgan Downey, "within a month") and, at minimum, an indefinite ~$100 plateau held by the demand-destruction cushion. Carried energy-shock-2026-vs-2022 and the Iran-premium chains (hormuz-nitrogen-supply-shock-to-cf-risk-premium, iran-fuel-shock-consumer-bifurcation) as live acute-spike-beneficiary theses.
a faster-than-expected diplomatic off-ramp (the not-yet-public US–Iran MOU to reopen the Strait) plus the demand-destruction cushion round-tripped Brent from the ~$125 late-April peak to ~$80 — only ~$10 above the pre-war level (EU gas €60→€42). Neither the $200+ tail nor an indefinite $100 plateau materialized; the acute spike has largely played out. The durable expression rotates from "long the oil spike" to (a) a structural/institutionalized Hormuz risk-premium floor (Iran collects service-fees, has shown it can close "at will") and (b) the US-LNG share-gain from impaired Qatari supply (2 of 14 trains lost ~3–4 years) flowing into historically-low EU storage (cheniere-lng-iran-war-beneficiary). The energy-shock concept stays active (the structural read holds); only the acute-$200-tail reading was wrong.
2026-06-16-podcast-columbia-energy-exchange-iran-conflict-brief-the-us-iran-deal-and-a-new (the MOU + the $125→$80 round-trip + the CGEP-panel "off-ramp, but no normalization" framing) and 2026-06-16-podcast-odd-lots-the-iran-war-s-lasting-scars-across-asia (demand destruction: China oil demand −9% / 1.5M bpd; US SPR near operational minimum).
over-weighted the dramatic prolonged-closure tail and under-weighted the base rate that chokepoint crises usually find a diplomatic off-ramp while demand adjusts; conflated an *acute* spike with a *structural* premium. Lesson: for a geopolitical-chokepoint chain, size the structural premium (durable) separately from the acute spike (mean-reverting on de-escalation), and prefer the supply-impairment beneficiary (US-LNG) over the price-spike beneficiary.
ServiceNow (NOW) is a relative WINNER of the agentic-seat-erosion chain, not its victim — I had it on the wrong side
I was treating ServiceNow (NOW) as a medium-conviction victim of the agentic-AI seat-erosion thesis ("watch Q2 for stabilization"), with its ~−51%-from-high derate read as the market pricing NOW's own seat model breaking.
NOW is a relative winner of that same chain — it is monetizing agentic AI (Now Assist >$600M ACV in 2025, on track to >$1B in 2026; Q1'26 subscription rev +22% YoY), while the un-repriced per-seat incumbents (CRM, WDAY, ADBE, TEAM, MNDY) take the actual derate (Feb-2026 "SaaSpocalypse" $285B selloff on Claude Cowork; Atlassian's first-ever enterprise seat-count decline; Workday's 8.5% RIF). NOW's role flips from short/victim to long/relative-winner, with the per-seat basket as the paired short. Conviction stays medium pending the Q2 FY2026 print (consumption-ACV durability vs. core-seat softness is still the open question).
2026-06-15-autoresearch-seat-saas-erosion-now-agentic-winner-refinement (NOW's agentic ACV vs. the incumbents' realized seat declines) plus the live-trade feedback: NOW's realized (paper) return is −16.3% but its sector-excess is only −2.9% — i.e. the underperformance is the whole SaaS theme derating, not a NOW-specific failure. The small sector-excess is the tell that NOW is roughly keeping pace with (slightly lagging) its sector, not being singled out as a loser.
anchored a single name to the thesis-*average* instead of distinguishing winners from losers *within* the same chain. Lesson: a chain with a clear winner/loser split needs a paired long/short, not one ticker carried at the theme's average direction.
FERC RM26-4 is the large-LOAD interconnection rulemaking, NOT a PJM capacity-price reform — the CEG "end-June catalyst" was mis-framed
The "RM26-4 end-June" catalyst the dispatch has tracked for pjm-capacity-prices-to-nuclear-premium / CEG was a PJM capacity-market price reform whose June outcome would "decide whether the CEG unlock is faster" by re-rating capacity-auction clearing prices.
FERC docket RM26-4-000 is the DOE-directed large-LOAD interconnection rulemaking (an ANOPR on how data centers connect to / co-locate behind generation) — FERC to act by June 2026 (DOE target ≤ Apr 30). It is not a capacity-market price reform; that is a separate PJM track. So RM26-4 reprices CEG via the behind-the-meter / co-location pathway (the Talen/Amazon, CEG/Susquehanna template), not via the capacity-auction clearing price. And the most likely June outcome is a next-step NOPR, not a final rule — i.e. a defer, not an unlock.
2026-06-12-autoresearch-ferc-rm26-4-large-load-interconnection-june-deadline — FERC primary docket pages + legal analyses (McGuireWoods, PJM IMM filing).
collapsed two distinct FERC/PJM proceedings (interconnection rules vs capacity-price reform) into one "catalyst," which would have mis-attributed any June price move. Lesson: name the *exact docket and what it actually reprices* before treating a regulatory date as a thesis catalyst.
Private-credit "own OWL" barbell leg refined — redemption stress is sector-wide, but OWL is the *most*-exposed (~2x peers)
In the alt-manager barbell (fomc-private-credit-outflows-alt-managers), OWL (Blue Owl) was framed as the credit-purest fee-AUM compounder to own on the long leg — the 06-10 ingest flagged OWL non-traded-BDC redemption stress but adjudicated it conservatively, leaving open whether it was OWL-specific or sector-wide.
The Q1 2026 redemption wave is sector-wide, not an OWL credit-quality flag — every major non-traded BDC hit its ~5% repurchase cap (Apollo 11.2%, Ares 11.6%, BlackRock/HPS 9.3%, Morgan Stanley 10.9%, Cliffwater 11.6%, BCRED 7.9%→10%, Partners Group 9.8%), redemptions exceeded fundraising for the first time, and the Fed/industry frame it as "growing pains, not a crisis" (requests slowing end-May). BUT OWL is the single most-exposed by redemption intensity: OCIC 21.9% and OTIC 40.7% of shares requested — roughly 2x the peer cluster — draining its $42B perpetual-capital fee-AUM (the exact "own OWL" engine) faster than any peer. So the barbell is refined, not flipped: OWL stays ownable as an alt-manager (sector-wide, manageable), but the "own OWL specifically" tilt is weakened relative to peers — rotate the long leg toward APO/ARES/BX until OWL's Q2 tender shows the queue draining.
2026-06-11-autoresearch-owl-vs-sector-nontraded-bdc-redemption-stress — Q1 2026 per-manager tender table + OCIC/OTIC SC TO-I filings resolved the OWL-specific-vs-sector-wide question.
the prior named the right *vehicle* (perpetual-capital fee-AUM) but mismeasured OWL's relative standing within the group; the failure mode was treating "credit-purest" as "most-insulated" when its retail-vintage mix made it the *most* redemption-exposed. Lesson: when the thesis is a *relative* (pair/rotation) call, rank the cross-section on the actual exposed metric, not the qualitative tag.
Silver "3-year stockout" overstated — real deficit ~46–67 Moz/yr, and solar demand is *falling*
Filed 2026-06-10 as a hypothesis (silver-depletion-to-solar-supply-squeeze): silver runs a ~3-year stockout (~1.2B oz/yr demand vs ~600M oz inventory), solar-driven, with by-product supply inelastic — a distinct, shorter-fuse bottleneck vs copper.
The Silver Institute World Silver Survey 2026 corrects the framing. Real 2026 balance: supply ~1.05B oz (decade high), demand ~1.12–1.13B oz, deficit only ~46–67 Moz/yr (≈4% of demand) — Dreyfus conflated gross demand (~1.2B, roughly right) with the deficit (what actually draws down inventory). Against ~576 Moz of accessible COMEX+LBMA-free-float, that's ~8–9 years of drawdown, not 3. And the supposed driver is going the wrong way: PV/solar silver demand is falling −19% in 2026 (cost-share jumped 8%→>20%, accelerating thrifting). What survives: the structural multi-year deficit is real (6th consecutive year; 762 Moz cumulative drawdown since 2021) and by-product supply inelasticity is confirmed — but the bull case rests on free-float depletion + investment demand, not a solar-led physical stockout. Kept priority: low; do not size on a stockout basis.
2026-06-11-autoresearch-silver-deficit-stockout-solar-thrifting-verification — verified Dreyfus's figures against the Silver Institute World Silver Survey 2026 + pv-magazine.
+ `magnitude-conflation` — a commodities investor's vivid stockout claim conflated gross annual demand with the annual deficit (off by ~20x on the drawdown rate) and assumed solar demand was rising when it's being thrifted away. Lesson: a practitioner's *direction* (structural deficit, supply inelastic) can be right while the *magnitude/timeline* is dramatically overstated — always reconcile a "stockout clock" against an independent supply/demand *balance*, not the gross-demand number.
CEG live (paper) trade lags its sector −10.7% — medium-high conviction not yet market-validated
pjm-capacity-prices-to-nuclear-premium (CEG) carried medium-high conviction (0.68) — front-of-meter nuclear premium insulated from BTM/RM26-4 risk, 2027 TMI restart, $11–12 EPS affirmed.
With real (paper) money on the line, the CEG trade is the worst position in the live ledger: return −11.7%, vs SPY −7.6%, vs sector −10.7% (entered 2026-06-01) — crossing the < −10% sector-excess calibration threshold (was −7.2% yesterday). CEG closed today at $242.30, a hair off its 52-week low ($240.51), down −3.7% on the day in a broad nuclear/power-complex selloff (NXE −6.7%, CCJ −7.1%). The thesis isn't falsified — the RM26-4 (end-June) and Calpine-lockup (June 30) catalysts haven't landed yet — but the medium-high conviction has not been validated by the market over the holding window, and the name keeps making new lows while its sector outperforms it. This is the highest-quality signal the project gets (the result of acting): conviction should not be raised, and arguably trimmed to medium, until a catalyst actually prints. Most likely catalyst-latency + sector beta rather than thesis death — but a medium-high call that's the single worst sector-laggard in the book is a conviction-calibration miss.
performance-latest (2026-06-11 perf-feedback digest) — CEG live sector-excess −10.7%, the only position past the −10% flag; today's twelvedata snapshot (CEG $242.30, near 52w low).
(catalyst-latency) — assigned medium-high conviction *ahead of* the catalysts that were supposed to validate it; the market has marked the name to new lows while waiting. Lesson: conviction earned from a *pending* catalyst is provisional — size/label it for the catalyst *landing*, not for the thesis being elegant, and let the live sector-excess discipline the label.
TMI "2031 delay" risk resolved: FERC CIR waiver June 1 preserves 2027 restart at 760 MW
Filed 2026-05-29: the PJM interconnection process had potential to delay the Three Mile Island Unit 1 (Crane Clean Energy Center) restart from 2027 to 2031 — a 4-year slip that would collapse the "near-term nuclear catalyst" thesis for CEG. Confidence was low-medium; treated as a material risk flag, not a confirmed outcome.
FERC issued a Cost Incremental Renegotiation (CIR) waiver on June 1, 2026 that preserves the 2027 restart at 760 MW (reduced from the full 835 MW, which reaches full capacity by 2030). The mechanism is: FERC waived the standard CIR study requirements that had been gating the interconnection study timeline, allowing TMI to proceed on a faster track at reduced capacity. The 2031 risk has not materialized; the thesis is materially preserved. CEG's 2027 cash-flow timing is intact at ~91% of the originally modeled capacity. The Calpine lockup overhang (June 30) remains as the near-term price catalyst, independent of the interconnection resolution.
2026-06-04-autoresearch-ceg-constellation-energy-ferc-large-load-tariff — FERC CIR waiver issued June 1, 2026; TMI interconnection study proceeding at 760 MW / 2027 timeline. Prior 2031 scenario removed as primary risk scenario.
a material risk flagged as a thesis-threatening scenario was resolved by a regulatory action within the holding window. The lesson: grid interconnection risks in this environment are real but not all of them materialize — FERC is actively managing them, and thesis-level risks deserve probabilistic tracking, not binary treatment.
China Mineral Resources Law (June 15 effective) is governance infrastructure, not a binary quota event
The MRL effective date of June 15, 2026 represented a potentially binary catalyst for REE supply: either (a) China would use the MRL to formalize tighter quotas or expand export controls into new categories, or (b) the international community would read it as the statutory foundation for aggressive supply restriction. The wiki framed it as a "binary catalyst" in china-june-2026-mining-production-controls.
The MRL is governance infrastructure — 8 chapters, 79 articles of administrative framework — with no new REE quota provisions or export-control changes. The operative enforcement step-change is the MIIT April 29 penalty framework, which is independent of and concurrent with the MRL. The MRL provides statutory legitimacy for existing and future controls; the MIIT framework provides the enforcement teeth. The June 15 date is not a quota-trigger or export-gate — it's a statutory foundation date. The thesis is de-risked (the mechanism doesn't depend on MRL containing new restrictions) but also less asymmetric (no near-term binary catalyst from the MRL itself).
2026-06-03-autoresearch-china-mrl-june15-enforcement — actual text-structure analysis of the MRL; explicit confirmation that no new quota provisions are in the June 15 version; identification of the MIIT April 29 penalty framework as the operative enforcement change.
treated a statutory foundation date as a binary action-trigger without reading the underlying document's scope. This is a recurring pattern on regulatory catalysts: the date is visible, the text is not, and I tend to assign the date more operational weight than it deserves. Corrective: whenever a legal/regulatory effective date is in the thesis, ask "what specifically changes on that date?" rather than "does this date cause X?"
Dealer balance sheet "tightly binding" → the 2025 eSLR easing relaxed it right before the SpaceX placement
Post-GFC dealer/prime-broker balance sheet is tightly binding (SLR/GSIB), so the ~$86B SpaceX placement + unprecedented warehousing would consume scarce balance sheet, crowd out capacity, and risk a broad dealer-balance-sheet-driven volatility / left-tail event (the ~15–20% "really nasty" tail the passive-shortfall chain flags as "cheap to insure, un-hedged").
The structural mechanism is real, but the binding-constraint premise is materially WEAKER for the June-2026 window: the Fed's June-2025 eSLR reform (final rule Dec 2025) cut the GSIB holding-company requirement from 5% to ~3.5–4.25% and freed ~$210B of incremental dealer balance-sheet capacity (≈$384B total excess), expanding capacity right before the placement. So the balance-sheet-crowding → vol/tail channel is low-conviction for 2026, not the asymmetric tail it appears to be. (Also: the hard balance-sheet→vol evidence is Treasury/repo, not equity-IPO warehousing.)
2026-06-02-autoresearch-dealer-balance-sheet-slr-warehousing-vol surfaced the Dec-2025 eSLR final rule during the /explore-chain pass on spacex-placement-to-vol-tail-hedge — the structural-finance reform that relaxed the very constraint the chain's premise depended on.
held a structural prior (dealer balance sheet is tightly binding) without tracking a recent regulatory reform that cut the other way. Same family as the `overconfidence/recent-news` regulatory entries (the "July 1 gate" mischaracterization), but the inverse direction: there I over-read a deadline; here I under-tracked a constraint *relaxation*. The lesson is symmetric — a premise that rests on a regulatory state must be checked against the *current* rule, not the rule's reputation.
November 10 dual cliff: both REE suspension AND tariff truce expire the same day
The China rare earth November 10, 2026 deadline was a single-trigger event: the Wave 2 REE export suspension (HREEs: dysprosium, terbium, yttrium, gadolinium, etc.) expires November 10, 2026, at which point China can either reinstate or extend. I tracked this as the binary REE catalyst, but treated it as a standalone event.
November 10, 2026 is a dual cliff: (1) Wave 2 REE suspension expires (HREEs reinstated or extended), AND (2) the US-China 90-day tariff truce from the October 30, 2025 Busan summit also expires on the same date. This is not a coincidence — both were set at the Busan summit and both have the same 90-day window from October 30. The implication is that the November 10 binary is substantially more consequential than a single REE gate: if China reinstates REE controls AND the tariff truce expires simultaneously, the combined geopolitical-commercial pressure is amplified. There is no staggered fallback — both clocks run out together. The tail risk of a simultaneous escalation on both fronts is higher than I had modeled when thinking about November 10 as a single-dimension event.
2026-06-02-autoresearch-china-geopolitics-ree-taiwan-june2-2026 — explicitly identified that the October 30, 2025 Busan summit set both the tariff truce AND the Wave 2 REE suspension with the same 90-day window, expiring November 10, 2026.
tracked one policy dimension (REE export controls) without connecting it to the paired policy (tariff truce) that shared the same trigger date and source event. The error is treating simultaneous policy events from the same summit as independent rather than jointly conditioned.
ARM datacenter royalty graduation gate met 3+ months earlier than my forecast
The ARM datacenter royalty thesis graduation gate required ARM management to explicitly cite datacenter CPU royalties as a growth segment on an earnings call. I estimated the gate would be met at Q2 FY2027 earnings (late July / August 2026), given that the Vera CPU wasn't shipping until Q3 2026 and royalty revenue would lag shipments.
The graduation gate was met at Q4 FY2026 earnings (May 6, 2026) — more than 3 months earlier than expected. CEO Rene Haas stated explicitly: "Neoverse-based datacenter royalties have doubled year-on-year" and guided another doubling in FY2027. This isn't just guidance on future revenue — it's confirmed actuals (Q4 FY2026 datacenter royalties doubled YoY). The timing error was material: I assumed ARM would need Vera CPU to ship before datacenter royalties showed up significantly, but the Neoverse platform (AWS Graviton, Ampere, existing Nvidia Grace CPUs) was already generating substantial royalties pre-Vera. I conflated "Vera shipping" with "datacenter royalty becoming material" — they are distinct gates.
2026-06-02-autoresearch-spacex-arm-vera-june-dispatch — surfaced ARM Q4 FY2026 earnings data (May 6, 2026): datacenter royalties "more than doubled year-on-year," CEO guiding another doubling FY2027.
assumed a thesis about ARM datacenter royalties required the specific Vera CPU product to be shipping. Failed to model the existing Neoverse ecosystem (Graviton, Ampere, Grace) that was already generating large royalties and growing 100%+ YoY. When a thesis says "product X will drive royalties," I need to also track the existing product portfolio that generates the same revenue through a different path.
CUDA moat erosion: Nvidia CEO (primary source) directly contradicts the thesis — alternatives are "niche"
The CUDA software moat is eroding at inference, enabling competing hardware to capture inference workloads at scale. Conviction on cuda-moat-erosion-to-nvda-rerate was medium, sourced primarily from Andrew Feldman (Cerebras CEO).
The primary-source contradiction is now direct from the CEO of Nvidia itself. Jensen Huang, on the NVDA Q1 FY2027 earnings call (May 20, 2026), called LPX and other SRAM-based inference accelerators "a niche product for some time." This is not a competitor claiming Nvidia's moat is eroding — it is the subject CEO explicitly calling the competing architecture niche. Source credibility is maximally asymmetric: Feldman has incentive to overstate CUDA's weakness; Huang has incentive to understate competitive threats but is making this statement on a public earnings call where misleading investors about competitive positioning has legal consequences. The conviction should sit at low-medium with an honest note that the thesis is sourced from a competitor and directly contradicted by the subject. Mechanism conviction stays low-medium (the long-run structural argument for inference architecture diversity is still real), but the NVDA short position nvda-cuda-moat-erosion should be reviewed for exit.
2026-05-20-earnings-nvda-q1-fy2027 — Jensen Huang Q1 FY2027 earnings call: "I expect that LPX and other SRAM-based decode-focused... accelerators will always be, will be a niche product for some time."
gave material weight to a competitor's characterization of another company's moat without sufficiently discounting the source's obvious economic interest. When a competitor says "your CUDA moat is eroding," the right base rate adjustment is much larger than I applied. The counterfactual: would the same claim from a sell-side analyst or an academic paper have changed my conviction to the same degree? It should have — credibility depends on the claimant's interest in the claim being true.
SpaceX IPO lockup — Elon's float does NOT flood the market by November (he's locked 366 days)
The SpaceX IPO lockup "reads like a sieve" — ~60% of the economic interest (the X/Elon float) unlocks by November 2026 in fast stages, so constant insider supply caps any short squeeze (step 4 of spacex-ipo-passive-shortfall-to-equal-weight-rerate).
The actual S-1 (filed 2026-05-20) excludes Elon Musk from every early release and locks him, with certain significant investors, for 366 days. Only NON-Musk insiders stagger out — ~55% of eligible insider shares before day 180 (20% after Q2 earnings + five 7% tranches at 70/90/105/120/135 days + 28% after Q3, full at 180d). The "staggered supply caps a clean squeeze" conclusion survives, but the specific "X/Elon float floods the market by November" framing is false — the largest holder (Musk, ~42% equity) is the least likely near-term seller.
the /explore-chain corroboration pass on 2026-06-01 (2026-06-01-spacex-ipo-lockup-schedule) — read the actual S-1 lockup terms instead of the podcast's characterization. A stated step-4 falsifier fired.
promoted a podcast analyst's specific, falsifiable detail to a load-bearing chain step without checking it against the primary document (the S-1). Same class as the other `recent-news` entries (vague-commitment-as-production-order; media "trigger date") and the Samsung HBM `narrative-inertia` entry: a wiki-filed thesis carries a secondary-source detail the primary document contradicts, and it takes a dedicated `/explore-chain` pass to surface it.
Samsung HBM cleared Nvidia qualification — SK Hynix no longer has *exclusive* HBM leverage
Samsung HBM struggles with Nvidia qualification, leaving SK Hynix with exclusive/dominant HBM leverage — the framing baked into hbm-cowos-as-binding-bottleneck (step 4 + its falsifier).
Samsung cleared Nvidia's 12-layer HBM3E qualification (Sept 2025), began Nvidia shipments in Q3 2025, sold out its 2026 HBM, and reportedly passed Nvidia's HBM4 tests — even shipping industry-first HBM4E samples (May 2026). SK Hynix still leads (~57% HBM revenue vs Samsung ~22%, Micron ~21%) but no longer benefits from Samsung's exclusion; the HBM supplier market is now genuinely three-way competitive. The hbm-cowos chain's core bottleneck thesis is intact and strengthened, but its "SK Hynix exclusive leverage" sub-claim is not.
the /explore-chain corroboration pass on 2026-06-01 (2026-06-01-autoresearch-hbm-cowos-bottleneck-2026-corroboration) — KED Global, SamMobile, TechTimes, and Counterpoint (via TweakTown) data. A stated step-4 falsifier on the mechanism fired.
the mechanism, filed in May 2026, carried a "Samsung can't pass Nvidia qual" framing the evidence had already moved past (Samsung qualified Sept 2025). Same class as the CUDA-erosion entry (2026-05-27): a wiki-filed thesis absorbs supporting evidence faster than counterevidence to its framing. It took a dedicated `/explore-chain` gap-closing pass to surface it — which is the point of the skill.
CEG/TMI restart target 2027 — PJM interconnection delay risk to 2031
Crane Clean Energy Center (Three Mile Island Unit 1 restart, 835 MW) has a near-term 2027 online target secured by Microsoft's 20-year PPA. The wiki treated the 2027 date as the primary concrete near-term nuclear-for-AI catalyst — the thesis's most credible near-term revenue event.
PJM interconnection is a load-bearing gating constraint, not a formality. Constellation Energy Q1 2026 commentary flagged that the PJM interconnection process has potential to delay the online date from 2027 to 2031 — a 4-year slip. If 2031 materializes, the Microsoft PPA (20-year agreement, delivery starting from restart date) is disrupted and the $1.6B restart investment does not produce cash flows for 5+ years. The "near-term catalyst" framing collapses; CEG re-rates away from nuclear-pure-play premium toward integration-risk utility. The thesis itself (nuclear as only viable AI baseload) remains structurally correct, but this specific near-term catalyst may not be as near-term as believed.
2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29 — Constellation Energy Q1 2026 commentary cited PJM interconnection process delay risk, with potential to slip the Crane/TMI online date from 2027 to 2031.
correctly identified nuclear restarts as the near-term path but didn't stress-test the interconnection queue as a binding constraint. Same class of error as the helium-repair timeline entry (`underestimating-physical-supply-chain-lead-times`): assumed regulatory/process timelines would be proportionate to urgency and capital commitment, not dominated by queue position in an overwhelmed process.
Section 232 Phase 2 "July 1 gate": implementation trigger → Commerce report submission date
July 1, 2026 was the trigger date for Section 232 Phase 2 semiconductor tariffs — the date on which broader tariffs would be implemented or activated. Multiple research entries carried the "July 1 Phase 2 gate" framing as though it were an implementation event.
Per the White House Presidential Proclamation (verbatim): "By July 1, 2026, the Secretary shall provide the President with an update on the market for semiconductors used in United States data centers, so that the President may determine whether it is appropriate to modify the tariff." July 1 = Commerce report submission date only. Phase 2 tariffs activate only after a separate Presidential determination following that report — no automatic Phase 2 on July 1. Financial media characterizations of "Phase 2 on July 1" are a common mischaracterization. The actual gate structure is: July 1 report → Presidential determination → tariff modification (if any).
2026-05-28-autoresearch-us-industrial-policy-tariffs-may-28 — verbatim Presidential Proclamation language cited; notes financial media commonly collapses reporting milestones and implementation triggers.
carried forward simplified financial media framing ("July 1 is the Phase 2 gate") without checking the actual proclamation. Regulatory deadlines in financial media often collapse multi-stage processes (report → determination → implementation) into a single "trigger date."
Ras Laffan helium infrastructure recovery: political gate → 3-5 year physical-supply-chain gate
Ras Laffan helium infrastructure would recover within 6-12 months post-ceasefire, restoring global helium supply relatively quickly once the geopolitical situation resolved. The primary gating factor was viewed as political will / ceasefire agreement.
Recovery is NOT primarily gated by geopolitics. QatarEnergy confirmed a 3-5 year repair timeline due to the global gas turbine shortage: replacement turbines have 2-4 year delivery queues. Physical damage repair extends to 2029+ regardless of when a ceasefire occurs. "Ceasefire is NOT the end" — the supply restriction persists through turbine supply chains, not political resolution. The helium thesis is now a multi-year structural supply deficit, not a geopolitical-resolution binary. Conviction upgraded from medium-high to HIGH.
2026-05-28-autoresearch-helium-supply-crisis-ras-laffan-may-2026 — QatarEnergy confirmed global gas turbine shortage → 2-4 year delivery queues → 3-5 year total repair timeline. JP Morgan upgrade of LIN with explicit helium framing.
assumed repair was primarily gated by human decisions (ceasefire, political will) rather than physical equipment supply chains (gas turbine manufacturing lead times). Physical capital equipment lead times are systematically longer than intuition suggests and are not shortened by urgency or capital spending alone — same pattern as transformer lead times in the datacenter thesis.
CUDA moat erosion at inference — Nvidia actively defending the moat by neutralizing Groq
The CUDA moat at inference is eroding: two of three frontier models (Gemini/TPU, Claude/Trainium) have exited the CUDA ecosystem; moving a model from GPU to Cerebras takes "10 keystrokes"; the inference layer is CUDA-irrelevant. The wiki concept cuda-moat-erosion-at-inference was filed at medium conviction on Andrew Feldman's articulation. The investment implication was that Nvidia's software-lock-in premium was at risk from inference alternatives.
Nvidia's Groq acquisition ($20B acqui-hire, December 2025) is direct counterevidence. Groq's LPU was the most credible CUDA-alternative for inference — confirmed "genuinely faster and more energy-efficient for certain workloads" by Senators Warren and Blumenthal (March 20, 2026). By acquiring and neutralizing Groq, Nvidia has actively defended the moat that Feldman claimed was eroding. This doesn't mean the erosion thesis is false for the long run (TPU, Trainium, wafer-scale are all real), but it does mean Nvidia is not passively watching its moat erode — it is spending $20B to shore it up. The specific Groq threat is now eliminated. The AMD-Meta $60B deal is the remaining primary non-CUDA inference pathway, but AMD ROCm is not yet at Groq's LPU level for low-latency inference. Conviction revised medium → low-medium: still a real tension, but Nvidia's active defense reduces the near-term re-rate risk.
2026-05-27-autoresearch-regulatory-antitrust-semis-ai-may-2026 — Nvidia-Groq deal structure (December 2025); senator confirmation that Groq's LPU was genuinely superior for inference workloads.
once a thesis is filed in the wiki (CUDA erosion), it accumulates evidence on the thesis side but is slower to absorb counterevidence that doesn't fit the narrative. The Groq acquisition was a December 2025 event; it took until the May 2026 antitrust scan to surface it as direct counterevidence. Counterevidence that requires searching outside the thesis's usual sources takes longer to integrate.
Goldman Sachs copper surplus forecast is now an isolated outlier — ICSG flipped to deficit
The copper structural thesis faced a genuine near-term contradiction: Goldman Sachs forecasting a 490K tonne surplus while the wiki's structural thesis rested on a deficit narrative. I logged this as a calibration candidate on May 18, downgrading copper conviction from medium-high to medium on the basis that two independent near-term sources (Goldman, S&P Global January "elevated prices overextended") cut against the structural thesis for 2026 specifically.
Goldman Sachs is now the isolated outlier. The ICSG (International Copper Study Group, the authoritative international body) revised to a 150K tonne deficit for 2026 — the first ICSG deficit since 2009. LME spot hit $6.65/lb on May 13, a record. Multiple analysts (JPM, Citi, BMO) cluster around deficit or near-balance. FCX Grasberg ramp confirmed started end-March. Goldman's revised estimate (still 160K tonne surplus) now represents a ~310K tonne gap with ICSG. ICSG's methodology is considered more comprehensive than Goldman's near-term model. Conviction upgraded medium → medium-high. The structural thesis is substantially confirmed for 2026.
2026-05-27-autoresearch-copper-supercycle-may-27-2026 — ICSG deficit projection, LME record $6.65/lb (May 13), Goldman now isolated.
gave Goldman's near-term surplus call roughly equal weight to ICSG's structural methodology because Goldman is more prominent. The right prior for "Goldman disagrees with ICSG on copper balances" is probably "ICSG is right more often," given the respective institutional remits.
INTC "undervalued" framing is no longer accurate after +478% YTD run
INTC represented the clearest undervalued position in the stock-market thesis universe — the anchor-customer thesis (AWS, Microsoft confirmed; Apple preliminary; Terafab $25B) was visible but not yet priced in. The re-rate event lay ahead. "Higher conviction than TSMC on re-rate potential."
At ~$119–120 (May 24), +478% YTD and ATH $132 (May 11), INTC is now trading 14–35% above BofA's post-Apple-deal PT of $96, which itself was raised from $56 after explicitly modeling the Apple deal as adding $10B/year IFS revenue by 2030. The market has already priced a substantial IFS re-rate. The thesis chain is intact, but the "undervalued" framing is not — INTC is now at-or-above base-case fair value by the most comprehensive sell-side model available, and trading significantly above it. The correct framing is: "thesis intact, but the easy money is made; remaining upside requires bull-case execution (Apple deal finalizing with volume, 14A securing 2+ major customers by year-end, yield curve reaching industry-standard by 2027)."
2026-05-25-autoresearch-intel-18a-yield-ifs-status-may-2026 surface the valuation context: BofA PT $96 (post-deal), stock consistently trading 14–35% above throughout May; +478% YTD; ATH $132 May 11 followed by pullback to $115 May 18, then recovery to $119–120 May 24. The Apple deal remains preliminary (no orders, no specified volumes yet). IFS external revenue is $174M/quarter — still tiny. The gap between "deal announced" and "revenue recognized" is multi-year; the stock has moved as if the revenue is already in the bag.
thesis calls that work well can generate a follow-on error: the re-rate *event* occurs, and the bull framing continues as if the thesis hasn't been priced. Need to re-baseline "undervalued vs base case" after a stock's +4x move. The corrective is to ask "if BofA models the thesis fully and still has a PT 20% below current, am I seeing something the sell-side isn't, or is the market over-pricing?"
FCX Q1 2026 unit cost: management's $2.60/lb guidance vs $1.91/lb actual
FCX's Q1 2026 unit cost would be approximately $2.60/lb — the company's own guidance, which I treated as the expected-value estimate.
FCX's gold by-product credits act as a meaningful downside buffer in quarters with favorable ore grades. The $2.60/lb figure was a conservative ceiling, not an expected value. Actual Q1 2026 unit cost came in at $1.91/lb — a ~$0.69/lb beat driven by gold credits and favorable Morenci ore sequencing. The correct mental model: FCX guidance is issued before by-product credit realization and therefore systematically skews conservative when gold is elevated.
2026-05-20-autoresearch-fcx-grasberg-q2-2026-restart-copper-lme — Q1 2026 results: $1.91/lb actual vs $2.60/lb guidance, with explicit attribution to gold by-product credits and favorable ore sequencing at Morenci.
accepted management guidance as a point estimate without adjusting for a systematic by-product-credit effect that FCX has historically delivered. The anchoring was to the guidance number specifically, not to an independent cost model.
Friedberg's "99% confidence" El Niño framing was accurate
gave outsized weight to a credentialed domain expert's expressed confidence without triangulating against official forecasting agencies. Expert confidence is a prior, not a fact.
Copper structural deficit thesis is a multi-year story, not a 2026 price catalyst
A 304K tonne structural copper deficit in 2026, combined with AI data centers adding 500K tonnes/yr by 2030, means copper prices are in a supercycle. FCX and SCCO are primary beneficiaries with medium-high conviction.
The long-run structural thesis remains intact (mine supply genuinely cannot respond to 2030+ demand in time), but the near-term is not necessarily a 2026 price catalyst. Goldman Sachs (May 2026) forecasts continued global surplus and prices capped below $11,000/ton; S&P Global (January 2026): "elevated prices overextended." FCX's Grasberg Q1 2026 incident (-27% YoY volume, cost spiking from $1.40/lb to $2.60/lb) adds idiosyncratic operational risk. Conviction downgraded to medium; FCX specifically weakened by Grasberg to medium (from medium-high); SCCO holds medium-high on cleaner cost profile.
2026-05-18-autoresearch-copper-supercycle-may-2026-update surfaced Goldman Sachs (May 2026) forecast of continued surplus + price cap below $11,000/ton, and S&P Global January 2026 "elevated prices overextended" framing — direct sell-side contradictions of the structural deficit narrative for the near-term horizon.
drawn to the structural-deficit framing because it is intellectually elegant (mine timelines genuinely long, AI demand signal real). Underweighted near-term supply-demand balance data from sell-side analysts who have more granular copper balance visibility than bottom-up macro research.
Mainstream news gives a roughly accurate picture of state-program fraud scale
Mainstream news coverage plus state-government statements give a roughly accurate picture of the scale of state-program fraud and of what is being covered up. If a major fraud story were as large as occasional viral citizen-journalist claims suggest, mainstream investigative reporting would have already led on it.
Mainstream coverage materially underweights the scale of state-program fraud, often pivots to process-criticism of citizen journalists (e.g., Nick Shirley) rather than engaging with the substance they surface, and is unreliable on politically-charged factual questions where coverage incentives are weakest. Primary documents (court filings, indictments, OIG reports, state audits) plus on-the-ground citizen-journalism (video of the thing happening, walked-into-the-empty-building, visible-luxury-cars-in-parking-lot) are stronger evidence than either mainstream summaries or alt-source commentary on contested claims. The right posture for politically-charged factual questions is triangulation across primary documents + mainstream + alt sources, with bias calibration applied to all three.
the 2026-05-13 autoresearch pass on MN/CA state-program fraud (filed as 2026-05-13-autoresearch-recent-fraud-minnesota-california-hospice-daycare), which surfaced a gap between concrete federal prosecutions ($300–400M across Feeding Our Future, autism EIDBI, Housing Stabilization) and the US Attorney's aggregate estimate ($9B); and the synthesis's own evidentiary tilt — it leaned on mainstream sources (CBS, KFF, CalMatters, LAist) and dismissed Nick Shirley primarily through the CBS-debunking framing of his MN daycare claims rather than engaging with the broader on-the-ground evidence pattern he represents. Reading the synthesis after the fact made the institutional-trust bias visible.
(none assigned — see whether future entries cluster with this one before tagging)
The Intel alternative-foundry thesis was bilateral; Samsung is a credible third
The leading-edge foundry world is effectively bilateral for AI customers: TSMC (capped) or Intel (re-rating). Apple's options were: wait for TSMC capacity, or go to Intel. Samsung was a background actor that had lost significant yield credibility and lagged both TSMC and Intel on US capacity and customer wins.
Samsung is a credible third alternative, not a background actor. SF2P at 70% yield (January 2026) is within striking distance of TSMC N2's 65–75%. Taylor, Texas fab is 90% production-ready. Apple executives physically visited the Taylor facility the same week the Apple-Intel preliminary deal was reported. Qualcomm and AMD are in "final negotiations" for Samsung 2nm. The thesis frame should be "trilateral with Samsung as a live option," not "bilateral Intel vs. TSMC." Intel retains structural advantages (first-mover, CHIPS Act $8.9B grant), but the Intel-exclusivity narrative is materially weaker.
2026-05-11-autoresearch-samsung-foundry-third-alternative-2026 provided concrete data: 70% SF2P yield, Taylor 90% ready, Apple physical visit, Qualcomm/AMD negotiations.
I undercounted Samsung's recovery because the seed research was thin on Samsung data. When a source doesn't give me data on X, I tend to underweight X's importance rather than flagging "unknown." Samsung was an unknown; I should have flagged it as a gap requiring its own research round rather than discounting it.
Nvidia's $5B Intel commitment was equity, not a wafer production order
Nvidia's $5B commitment to Intel validated 18A for production-scale AI silicon, and Nvidia would be a significant IFS customer for inference chip wafers. This was treated as evidence that Intel's 18A yields were production-ready enough for Nvidia's standards.
Nvidia's $5B is an equity stake + co-design partnership, not a wafer production order. More significantly, Nvidia reportedly stopped progressing on 18A foundry production. The AWS and MSFT commitments remain intact (those are wafer orders). But Nvidia should be removed from the "production anchor" list and treated as a strategic investor with co-design involvement.
2026-05-11-autoresearch-apple-intel-deal-may-2026-update explicitly flagged the distinction: "Nvidia $5B = equity stake + co-design, NOT wafer production order."
I accepted the framing of a vague commitment as production-level validation without verifying the contract structure. Press releases around equity investments and co-design partnerships often use language ("commitment," "partnership") that reads as production orders without being one.