Deep Dives

BITB — Bitwise Bitcoin ETF · Deep Dive

Position: 19,400 sh ≈ $683K · 22.6% of NAV (largest position) · Sleeve: Bitcoin · Role: ANCHOR

Written: 2026-07-28 · Next review: 2027-01-31 or on any falsifier event

1. Thesis

Bitcoin is in the bear phase of its fourth four-year cycle, and this bear is a bottoming

process, not a secular top. The structural case: spot-ETF plumbing survived its first full

drawdown intact; long-term holders own a record ~16.1M BTC; the largest single whale

accumulation spike on record printed near $59K; and the book's own power-law model reads

deep value (composite ≈ −1.1, price ≈ 0.5x the ~$117K fair-value line, Mayer 0.79,

Fear & Greed in the teens). The bet is that the 2026 lows are within the accumulation zone

of a cycle whose expansion phase runs 2027–28.

2. Vehicle choice

BITB over the alternatives, deliberately: spot exposure at a 0.20% fee with zero leverage,

zero credit risk, zero key-man risk. MSTR is a leveraged wrapper (owned indirectly via STRC

on the income side instead — the two legs are intentionally separated); futures ETFs bleed

roll; self-custody adds operational risk the size does not justify. BITB is exempt from the

R2 single-name cap by design because it is the asset, not a company.

3. Edge and base rates

The claimed edge is behavioral and structural, not informational: willingness to hold and

accumulate through a −50%+ drawdown that flows-driven sellers (ETF outflows $200–500M/day

at the worst; IBIT sold ~$5B over 60 days) cannot. Base rates that frame the bet:

• Every prior cycle put its bear-market low 12–18 months after the peak; this bear is

~9–10 months old (peak Sep–Oct 2025). Cowen's midterm-year frame: bears run ~50–60

weeks, modal low around October 2026.

• Prior cycle bears bottomed at −77% and −84% from peak; this one has run ~−55% from

~$126K to the $57–60K zone — consistent with the milder-bear/institutional-era view,

but the low-$30Ks tail (Wadsworth) is not dead and must stay priced.

• Post-halving cycle expansion has followed every accumulation zone so far (n=3 — a

small sample; that smallness is the honest core risk of the whole thesis).

4. Key numbers (as written)

BTC ~$59–66K range for six weeks. Model: fair value $117K, composite −1.07 to −1.29

(lump-sum trigger −1.3), 200DMA ~$77K (0.85x), Feb low $62,791 broken in June (bearish

structure), double-bottom attempt at $59–60K. 89% of supply unmoved in 3 months

(2nd-highest illiquidity ever). Position cost basis ≈ $33.69/sh vs $35.2 last mark:

roughly flat on a position built across the cycle.

5. Falsifiers and pre-committed actions

• Cycle failure: no new all-time high by end-2027 → the four-year frame is broken;

cut the anchor to core size (≤15%) regardless of narrative.

• Structural break: US regulatory hostility to spot ETFs, or a sustained ETF outflow

regime that persists even through rate cuts → thesis void, exit to target.

• Model break: power-law fair value revised materially down by the model's own data.
• Adds: ONLY via the pre-committed triggers (btc_repair weekly-close, btc_lumpsum at

composite ≤ −1.3). No narrative adds. Cluster cap: digital assets ≤35% NAV binds the

total (BITB+BSOL+COIN currently ~27.7%; STRC counted separately as income).

• Governor override: while the drawdown governor is at rung 3, no adds at all, even

if a trigger fires — capital preservation outranks accumulation.

6. Horizon

Full cycle: hold through 2027–28 expansion. This is explicitly not a trade; the SRS/

monthly-chart discipline applies to adds, not to the core.

7. Verdict

KEEP at current size. The position passes: it is the book's largest conviction with a

documented model edge, statement-verified sizing inside the cluster cap, pre-committed

falsifiers, and a defined horizon. What it does NOT justify: adding while the governor is

RED, or letting the digital-assets cluster creep past 35% via BSOL/COIN drift. The single

largest intellectual risk is sample size — three prior cycles is weak evidence, and the

whole anchor rests on it. Size (22.6%) is aggressive but consistent with the ratified

allocation target; revisit at the January 2027 review or a falsifier, whichever first.

TSLA — Tesla · Deep Dive

Position: 1,500 sh ≈ $464K · 15.3% of NAV (2nd largest; ABOVE the 12% hard cap) · Sleeve: AI · Role: CORE

Written: 2026-07-28, on a 108-agent verified evidence pack (25 sources fetched, 122 claims extracted,

25 adversarially verified 3-vote each: 24 confirmed, 1 refuted) · Next review: on Q3-26 earnings or any falsifier

1. Thesis (as held)

Robotaxi/FSD commercialization inflecting, energy storage as second engine, market pricing it as a carmaker.

The deep dive's job: test that against primary evidence and price the pieces ourselves.

2. What the verified evidence says

The good (primary-source confirmed):

• Q2-26 revenue $28.2B, +26% y/y; record 480,126 deliveries (+25%, ~1.9M annualized). Demand is not the problem. [Tesla IR Q2-26 deck + deliveries release]
• FSD: 1.48M active subscriptions, +56% y/y, >55% North America attach on new deliveries, five new EU

country approvals. This is the strongest single piece of bull evidence — a real, growing, high-margin

software stream. [Tesla IR Q2-26 deck]

• Energy deployments 13.5 GWh in Q2, +41% y/y. Volume growth intact. [Tesla IR]
• Cybercab production started at Giga Texas, >125K/yr installed capacity. The dedicated-vehicle path exists. [Tesla IR]

The bad (primary-source confirmed):

• Margins collapsed: Q2-26 GAAP operating margin 1.4% (-57% y/y op income); TTM net margin ~3.7%

company-wide ($3.8B NI on $103.6B revenue). The volume came from price cuts. [Tesla IR; arithmetic verified]

• Energy REVENUE grows only +13% (TTM $12.8B) despite +41% GWh — $/kWh deflation eats the volume; gross

margin compressed to ~20.4% by vendor-cell warranty charges. [Tesla IR]

• Robotaxi scale is trivial: ~2.5M cumulative paid miles ever (IR deck chart). Texas DMV registry: 42

authorized Tesla vehicles vs 577 Waymo (May-26); ~16-20 active unsupervised in Austin, shrinking since

April. Bloomberg counts ~59 registered total. [TX DMV via multiple outlets; crowdsourced trackers]

• Remote teleoperators still DRIVE the vehicles (sub-10mph authority per Tesla's statement to lawmakers);

three NHTSA-filed collisions occurred while a teleoperator was driving. Austin service availability was

~19% of operating hours as of Feb-26. [NHTSA SGO filings, unredacted May-26; Robotaxi Tracker]

• Federal overhang: NHTSA Engineering Analysis EA26002 (Mar-26) covers ~3.2M FSD vehicles on camera-only

degraded-visibility performance. Open as of July. [NHTSA ODI resume, primary]

• The benchmark that reprices the dream: Waymo runs ~3,000-3,600 driverless vehicles doing 500K+ paid

rides/week — roughly 100x Tesla's robotaxi scale — and books only ~$100-130M revenue per QUARTER.

Even the category leader at scale is a sub-$1B/yr revenue business today. [NHTSA recall scope; Waymo

announcements; FutureSearch estimates — the unit-economics numbers are estimates, flagged as such]

Refuted in verification (0-3): the claim that the service ran in only four cities with a 10-14 sq-mi

Miami zone — the IR deck's seven-metro list is the anchor.

Evidence gaps (honest): no verified robotaxi unit economics for anyone (Tesla's cost/mile, pricing,

utilization; Waymo's profitability); no verified sell-side implied-robotaxi decomposition; China FSD and

Chinese competitor scale unverified. The $40K-gross-profit-per-vehicle-year input remains judgment.

3. Our model (evidence-revised 2026-07-28)

Sum of the parts, 3.95B fully diluted shares:

| Component | Inputs | Value/sh |

|---|---|---|

| Auto hardware | 80B rev × 4% normalized margin × 25x | ~$20 |

| FSD software | FY28E 3.2M subs → 3.3B rev × 65% × 35x | ~$19 |

| Energy | FY27E 15B rev × 8% × 28x | ~$9 |

| Robotaxi bear (45%) | program stays niche; PV $20B | $5 × 0.45 |

| Robotaxi base (40%) | Cybercab ramps to 300-500K vehicles by ~2030-32; PV $300B | $76 × 0.40 |

| Robotaxi bull (15%) | 1M+ global fleet + licensing; PV $1.2T | $304 × 0.15 |

| Fair value | | ~$126 |

vs price ~$306 → our model says -59%. Street consensus $435 (cross-check only, never an input).

Probability sensitivity (the whole argument lives here):

| Probability set (bear/base/bull) | Option value/sh | Fair value |

|---|---|---|

| Evidence-weighted (45/40/15) — ours | $78 | ~$126 |

| v0 seed (35/45/20) | $97 | ~$145 |

| Bull-tilted (25/45/30) | $127 | ~$175 |

| Required to justify $306 | ~$258 | $306 |

Market-implied expectations: at $306 the market pays ~$1.02T for the autonomy option (price minus

core). That is ~85% of our BULL-case PV — i.e., the market prices near-certain success of a program that

today has ~50 running cars, teleoperators doing the driving, and a category leader 100x its size booking

under $1B/yr of revenue. For OUR model to be wrong and the market right, Cybercab must ramp near its full

125K/yr capacity immediately AND unit economics must exceed anything any operator has demonstrated.

4. Falsifiers — both directions, pre-committed

Against the position (act on these): EA26002 escalates to recall of camera-only FSD; teleoperator

dependency persists into 2027; Cybercab run-rate <50K/yr by mid-2027; FSD subscription growth <25% y/y.

Against our bearish model (raise the PVs if these print): driver-out fleet >1,000 vehicles across 5+

metros; disclosed robotaxi revenue run-rate >$500M/yr; supervision-free miles-per-intervention data at

human-parity; FSD subs >2.5M by mid-2027 (raises both FSD leg and base-case probability).

5. Horizon

The autonomy option resolves substantially by 2028 (Cybercab ramp + EA26002 + driver-out data). The

position is NOT a hold-forever compounder at this price; it is an option trade where we now have evidence

the option is expensively priced.

6. Verdict

TRIM to the 10% target (R2 rule), keep the remainder. The thesis kernel is real — FSD software growth

is the best-verified bull fact in the book, and the Cybercab line exists. But our own model, built on

primary sources, says the current price embeds ~85% of the bull case while the operational evidence sits

closest to the bear case. Holding 15.3% (above the hard cap, 12 days in breach) means betting overweight

AND against our own model AND against the governor's rung-3 de-gross order. Trimming ~$160K to 10%

satisfies the R2 breach, funds the cash floor the governor demands, and keeps full upside participation

on two-thirds of the position if the FSD evidence keeps compounding. Conviction: 4/10 at this price

(would be 7/10 at ~$150). Evidence grade: HIGH (verified pack on file).

7. Challenge model: Cern Basher (added 2026-07-28, Ian's request)

Basher joined the ingestion roster as the named TSLA challenge model. A 95-agent verified research pass

reconstructed his published record (X posts date-verified via tweet-ID decoding; tables OCR'd):

• His most recent full model (Aug 25, 2025 Sum-of-the-Parts): deliveries 1.7M → 7.5M by 2030 at 20%

net margins (auto worth up to $900B); energy to 3,000 GWh by 2035 (up to $1.35T); robotaxi ~10M

vehicles by 2030 at $90/day/vehicle; Optimus 1.55M bots by 2030 at $135/day. Full stack: $2,452–$5,425/sh.

• His own bear anchors: core-business-only floor $132–$234/sh; halved-volume auto case ~$450B;

spoken range (Mar-25) "probably between 200 and a thousand." Every table footnoted

"Illustration, not a prediction," with no published probability weights.

• No verified 2026 model revision after actual robotaxi fleet data, Q2-26 earnings, or the stock's

2026 decline. His only verified 2026 output is a July FSD chart: ~666K paying subscribers, ~$800M ARR.

Resolution of the input disputes (evidence-settled): his 7.5M-by-2030 deliveries need ~41%/yr

compounding vs the verified 25%; his 20% auto net margin has never been achieved (verified TTM 3.7%,

historic peak ~15%); his 10M-robotaxi-2030 fleet requires ~80x the Cybercab line's installed capacity.

None of these inputs move our model, because none carry evidence. One of his data points DOES move us,

in the bearish direction: his ~$800M FSD ARR chart implies our FY28E $3.3B FSD revenue leg is likely

too high — flagged for the next assumptions revision.

The verdict this produces: adopting "Basher's bear case as our base case" — the literal request —

sets fair value at $132–$234, bracketing our $126 and sitting 24–57% BELOW the market price. His

divergence from us is not about the business that exists; it is about assigning near-certain success

to a 10M-vehicle robotaxi fleet and an Optimus empire, without probabilities. Our model prices those

as options with evidence-weighted probabilities. The TRIM-to-10% verdict stands, reinforced.

PSCE — Invesco S&P SmallCap Energy ETF · Deep Dive

Position: ~$70K · 2.3% of NAV · Sleeve: Energy · Role: SATELLITE

Written: 2026-07-29 · Next review: n/a if swap executes

1. Thesis (as held) vs what the vehicle actually does

The thesis card said: "small-cap E&P basket, cheap on cash flow, levered to the oil geopolitical tail

without single-well risk." The deep dive tested the second claim with our own 895-day study and it is

false in the way that matters: PSCE's spike-day capture of Brent moves >+2% is 0.48 — the tail

exposure is half of what the thesis assumed. On 2026-07-28, with Brent +3.8% on the repelled Iran

attack, PSCE printed +1.4% while XOP did +3.0% and BNO ~+3.5%. The vehicle is fine; it is simply not

the instrument for the thesis it was bought to express.

2. What PSCE actually is

Small-cap S&P 600 energy: junior E&Ps plus oilfield services and equipment. Its returns carry small-cap

equity beta (drag on red tape days), heavy producer hedging (mutes near-term price capture), and a

services component with no direct crude linkage. It IS a reasonable multi-year value vehicle: +35.4%

trailing 12 months, and it beat spot Brent over that window. But the book holds it against a dated

spike thesis (thesis 3, 38% conviction, post-midterm window), and for that job it is mis-specified.

3. Verdict

SWAP OUT into BNO, dollar-for-dollar (see deep_dives/BNO.md for the receiving thesis, capture

table, EV gate, and pre-committed exits). No net risk change; governor respected. If Ian would rather

own the VALUE thesis than the SPIKE thesis, the correct alternative is keeping energy-equity exposure

via XOP (broader, purer E&P) — but then the position should be re-labeled as a valuation bet with a

multi-year horizon and removed from the thesis-3 expression list, because it cannot do that job.

Evidence grade: HIGH (own capture study). Status: REVIEWED.

BNO — United States Brent Oil Fund · Deep Dive (pre-trade)

Proposed position: SWAP from PSCE, same dollars (~$70K, ~2.3% of NAV) · Sleeve: Energy · Role: SATELLITE (tail expression)

Written: 2026-07-29, pre-trade per the no-position-without-REVIEWED-thesis rule · Next review: thesis-3 conviction change, or 2026-11-30 (post-midterms), whichever first

1. Thesis being expressed

Thesis 3, the oil geopolitical tail: the Iran war resumes or Hormuz risk reprices (Pape framework: >50%

odds of renewed full war after the midterms; today's US-repelled Iran attack on a base is the tail firing).

This is a PRICE-SPIKE bet with a dated window, not an energy-equity valuation bet. The instrument must

capture a fast, front-month, seaborne-crude repricing.

2. Vehicle selection — the whole point of this dive

Our capture analysis, 124 Brent spike days (>+2%) since 2023, ~895 trading days:

| Vehicle | Spike-day capture | Beta to Brent | Corr | 12m return vs Brent +20.2% |

|---|---|---|---|---|

| USO (WTI futures) | 0.96 | 0.89 | 0.90 | +61.7% |

| BNO (Brent futures) | 0.92 | 0.86 | 0.92 | +55.7% |

| XOP (E&P equities) | 0.50 | 0.49 | 0.62 | +32.8% |

| PSCE (small-cap energy) | 0.48 | 0.48 | 0.55 | +35.4% |

| XLE (large-cap energy) | 0.37 | 0.36 | 0.59 | — |

Why BNO over USO despite USO's slightly higher historical capture: the scenario is a SEABORNE supply

shock. Roughly a fifth of global oil transits Hormuz; Brent is the waterborne global benchmark that

reprices it directly. WTI is the landlocked US benchmark buffered by Cushing storage, domestic supply,

and SPR politics — in Gulf-specific shocks the Brent-WTI spread widens, and USO's capture of that

specific scenario degrades exactly when the thesis pays. USO's 0.96 average is earned across ALL spike

days (many macro/demand-driven, where WTI keeps pace); BNO's capture is scenario-robust. We pay ~4pp of

average capture for fidelity to the actual bet. Statistically close; mechanically clear.

Why not stay in equities (XOP): half the capture (0.50), plus equity-market beta the governor is

trying to shed. XOP is the right vehicle for a multi-year energy VALUE thesis — that is not thesis 3.

3. Carry: currently a tailwind, watch for the flip

Both futures funds beat spot Brent by 35-41pp over the trailing 12 months — the war regime kept the

curve in backwardation, so rolling futures EARNED carry. Caveat, honestly: the June MOU + IEA glut

forecast can flip the curve to contango, turning carry into a bleed of roughly 5-10%/yr. That is the

cost of holding the tail. If the tail closes, the position exits anyway (falsifiers below), so the

worst-case is bounded bleed over a bounded window.

4. EV gate (run before sizing, per the RUNBOOK)

U (Hormuz/war event): Brent $110-130 (+30-50%) x 0.9 capture → +27-45%; use +0.35.

D (durable peace + glut): Brent to ~$70 (-20%) plus contango bleed → -0.25 over the horizon.

Breakeven p = 0.25/(0.35+0.25) = 0.42. Honest p = thesis-3 conviction = 0.38.

The gate says: below breakeven as a NEW bet — do not add. As a SWAP it still clears, because the

book already carries this thesis in a vehicle with HALF the upside capture and the same downside class:

swapping dollar-for-dollar strictly improves the payoff profile of risk already held. Sizing therefore

stays at PSCE's current ~2.3% — no increase, governor stays respected (swap, not add).

5. Falsifiers and pre-committed actions

• Thesis-3 conviction drops below 25% (durable peace holds through the midterm window) → EXIT, no debate.
• Brent closes below $70 → the glut regime won; EXIT.
• Curve in contango steeper than ~8%/yr annualized for 2+ months with no escalation → the bleed

outruns the tail; EXIT or halve.

• Event FIRES (Brent >$110): take the spike — trim at least half into strength per the spike-selling

discipline; this is a tail harvest, not a new trend thesis.

6. Horizon

Dated: through the post-midterm window (thesis 3's own framework), review 2026-11-30 hard stop.

This position is not a compounder; it has an expiry mindset.

7. Verdict

SWAP PSCE → BNO, same dollars (~2.3% NAV), no size increase. Today's tape was the proof: Brent

+3.8%, PSCE +1.4%. If the tail fires with the book positioned in PSCE, we get half-paid for full risk.

Trade is Ian's to execute; this entry satisfies the pre-trade thesis rule. Conviction 5/10 (bounded by

thesis 3's 38%); evidence HIGH on the vehicle mechanics (our own 895-day capture study, on file).