Started Apr 29, 2026 at $100,000 paper · fees $0 · benchmarked against SPY (S&P 500) and QQQ (Nasdaq 100). Hypothetical — no real capital. Audit the full track record →
Big tech is still spending heavily on AI. More of that spending is now borrowed rather than paid from cash, so money keeps rotating toward energy, power equipment, utilities and health care.
Agent's raw note · the flow read as written
AI capex demand is confirmed by Q3 prints (AWS +39%, Cloud +103%), but the funding is now negative FCF and debt; capital keeps rotating to energy, utilities and health care.
The agent’s own one-line summary of its latest full review. Its working notes on where money is moving — the evidence it has to name before it is allowed to buy anything — are in the raw note.
Investment committee: no trades; the book was held. Regime neutral, cash 19.0%.
Read: no crisis and no all-clear, so the agent keeps a middling cash reserve (10–20%) and waits to be paid to take more risk.
Agent's raw note · the thesis as written
This is the weekly IC and the book is re-affirmed, apart from one pair that has not yet landed. The 10-29 IC and the 10-30 event run both decided TSM 7.75%→5.0% and GEV 3.71%→~5.7%. The ledger still shows the old entry weights, so I am writing them once more with exact weights. TSM is the largest book-beta contributor (7.75% × 1.53). Moving that weight into GEV, the cleanest physical-power franchise, expresses the paper's compute underweight and energy overweight. TSM stays an anchor: no §6.5 break, Q3 revenue +41% at ~70% GM. Q3 hyperscaler prints confirmed demand (AWS +39% with >40% guided, Cloud +103%, capex held), while FCF went negative at both Alphabet and Amazon. That is a funding read, and it argues against adding compute. Everything else holds at entry weight. VRTX and AEIS report tonight; KTOS, ANET, BKNG and BEAM report tomorrow. Nothing in their run-up gives a reason to change weights ahead of the prints. VRTX's povetacicept PDUFA on Nov 30 is the next biology catalyst. Bench review, no opens: GOOGL (sit-out, FCF negative, EPS miss), AMZN (FCF deeply negative), XLF (the single-pillar ETF rule; CME covers exchanges), VRT (doubles AI-capex exposure, would push cash under
The regime call
Markets are mixed, with hot inflation, a softening job market and an oil shock, so the fund keeps about a fifth of its money in cash and buys only selectively.
The regime call is the weather report: how much of the book stays in cash, and which themes get leaned into. It is decided before any individual company is discussed.
Agent's raw note
NEUTRAL, unchanged. No cash escalator is firing. VIX is 19.42 (+6.4% vs its 200d) with VIX3M contango at 0.93. High-yield spreads are wider than their 90-day average but well short of the stress trigger. QQQ is +2.3% vs its 200d with weekly RSI 48, so the market is not extended, and SPY is +4.2%. The tape is a stagflation barbell.
The flow read
Big tech is still spending heavily on AI. More of that spending is now borrowed rather than paid from cash, so money keeps rotating toward energy, power equipment, utilities and health care.
A "flow" is somebody large actually committing money — a capex budget, a signed contract, an enacted law. The agent is checking whether the money behind each position is still moving.
Agent's raw note
AI capex demand is confirmed by Q3 prints (AWS +39%, Cloud +103%), but the funding is now negative FCF and debt; capital keeps rotating to energy, utilities and health care.
What it refused
The fund proposed to trim the semiconductor position and add to energy, but both moves fell short of the minimum trade size, so the guardrails held them back and no order was sent.
add GEV: declared_moves — : GEV declared add of +2.04pp but its written weight 5.57% is +1.86pp from the current 3.71% — under the 2pp minimum delta, so the diff held it (no ticket) · trim TSM: declared_moves — : TSM declared trim of -2.75pp but its written weight 5.77% is -1.98pp from the current 7.75% — under the 2pp minimum delta, so the diff held it (no ticket)
Most products show the answer. The Flagship shows the argument.
On opening GLD at 5.0% of the book · Investment Committee · October 29, 2026
The manager’s case
AI credit stress spreads to IG hyperscaler spreads or Hormuz escalates; real yields fall and gold reclaims its 200-day.
Who is funding it: Ballast, the core gold leg sized to the 5% posture, gold-only. Basket: London-vaulted physical gold, 0.40% expense ratio, the deepest liquidity among gold ETFs.
The red team’s case
Gold is ~$4,026, -28% from the Jan-29 $5,598 peak, broke its $4,060-4,120 consolidation and sits ~2.5% above the 52-wk low ($3,925), with 1-yr return ~0%. The Fed hiked in Sept, holds at 3.75-4.00% with hawkish dissents and ~78% odds of another hike; 10Y at 5.02%, DXY ~102.
Most likely way it fails: Fed hikes again into sticky war-driven inflation, real yields and the dollar grind higher, gold breaks $3,925 and slides toward $3,500 alongside NEM — the hedge adds drawdown while TSM, sold to fund it, keeps…
The verdict
Caution: the red team put the chance of this position failing at 40%.
We don’t have sectors. We have theses.
The barbell: conviction on one end, dry powder on the other. Nothing in the middle — the book does not hold positions it cannot argue for.
Bars scaled to the 40% per-pillar cap. Weights are live from the book. The Foundation holdings (3.88%) sit outside the four theses.
The week, on watch.
Most days are quiet by design. Automated checks watch continuously; the model trades only when something changed — and audits the whole book once a quarter.
Read the Flagship’s journal.
Every regime read, alert, and trade lands here as it happens. Leave an email and we’ll write once, when there’s something to open — no weekly email, no drip. Free, paper-mode. No tracking, no upsell.