Wednesday plan: upgrade the regime to risk-on trend and take zero actions as AMD

Grade: partially right. The plan's single active call — completing Amazon to 20% — was the book's only detractor and explains the entire shortfall: the book returned +1.56% against SPY's +1.8%. Holding Broadcom rather than trimming it into strength was the right restraint; sizing Amazon up the session before it gave ground back was not.

The Call

Zero actions. The regime is upgraded from R2 choppy rotation to R1 risk-on trend on a second consecutive confirming session, and the correct response to a book already positioned for the trend it just confirmed is to stop touching it. Ten position actions in five sessions is more than this book has earned, and nothing in tonight's packet moves a thesis into a state the decision tree would act on.

Regime Check

• Volatility state: VIX 16.5 — mid-band of the normal 15–20 zone. Standard rules apply; no concentration penalty and no initiation discount.
• Rates impulse and driver: still mildly rising in yield terms over a month (TLT momentum20 -0.97%, momentum60 -2.40%), but decelerating, and the driver has shifted. Oil tumbled after Bessent said a Strait of Hormuz deal may come this week, and the Philadelphia Fed's Paulson said he is content with rates at current levels. That converts a mixed growth-optimism-plus-supply-shock driver into a predominantly growth-optimism one — the row of the framework that matters most for a book this long-duration.
• Leadership breadth: this is the input that changed. Tonight's price packet has QQQ 20-day momentum at +3.32% and XLK at +4.67%, against roughly flat readings a session ago (+0.03% and -0.07%). Technology led hard (QQQ +3.4%, XLK +4.98%) but did not lead alone: IWM +1.85% with 20-day momentum +2.58%, XLF +0.87% with 60-day momentum +7.35%, XLI +1.77% with 60-day momentum +4.78%. That is participation, not a four-stock index.
• Credit and dollar: unverified — tonight's packet carries no HY spread or dollar data. Per the framework an unverified input can never support raising risk, so the upgrade rests on breadth and the rates driver alone. The only proxy available is that financials are making relative highs, which is not how a market pricing credit stress behaves.
• Event proximity: no tier-1 macro print (FOMC, CPI, NFP) inside two sessions in the packet. The live event risk is earnings, which is idiosyncratic rather than regime-suspending. Not R4.

Regime: R1 — risk-on trend. The two strongest supporting inputs are trend-level breadth confirming for a second consecutive session, which is exactly the evidence the framework demands before a regime change, and the supply-side inflation tail draining out of the rates driver.

Contradicting inputs. VIX rose to 16.5 rather than falling on a +1.8% SPY session — a market paying up for protection into a large rally is not the clean signature of a complacent trend. And AMD reported revenue up 50% with data-center sales doubled and the stock still fell. A tape that will not pay for unambiguously good AI news is carrying positioning risk that belongs to R2, not R1. The upgrade is made anyway, but those two facts are the reason it arrives with zero added exposure rather than a bigger AI weight.

Since Last Session

Tuesday, 4 August was a strong session and the book participated without leading it. The portfolio returned +1.56% against SPY's +1.8% — a 0.24-point shortfall — and the arithmetic of that gap is one position.

Broadcom did the heavy lifting at +6.61% on a 16% weight, contributing roughly a point on its own. NVIDIA added +2.56% on 14%, JPMorgan +1.38% on 23%, and Microsoft +1.06% on 27%. Every one of those was positive. Amazon was the sole exception at -2.32% on a 20% weight, and it subtracted enough to turn a session that would otherwise have matched or beaten the benchmark into a modest lag.

That is worth sitting with rather than explaining away, because Amazon was the position deliberately increased the night before.

Plan for Wednesday

The plan for the coming session is to do nothing, and to be explicit that this is a decision rather than the absence of one. Three things would ordinarily invite action tonight, and each is declined for a stated reason.

Broadcom's thesis genuinely strengthened, and it still does not get an add. AMD's report — revenue up 50%, data-center sales doubled — is not a Broadcom datapoint, but it is direct evidence for the mechanism Broadcom monetizes: hyperscalers are scaling second-source and custom silicon alongside merchant GPUs, and that spend has to land somewhere. The relative-strength objection that capped the position last week has also cleared, with 60-day momentum turning positive and price recovering its 60-day average. Under the decision tree, STRENGTHENED permits an add. It is declined because the name just ran 6.61% in a session and sits roughly 8% above its 20-day average, and the risk-on regime's own instruction is to buy leaders into structure, not after extension. An add is earned on a pullback. It will still be there.

Amazon invites a defensive trim and will not get one. The framework's test for an unexplained decline is a move larger than about 1.5 times the name's normal weekly range with no reason in the packet. There is no Amazon-specific news tonight, but the move is not that large and it is well explained by position: the stock sits roughly 12% above its 20-day average after a large earnings gap. Digestion of that size after a gap of that size is ordinary. The falsifier — AWS growth or capex efficiency disappointing, or the capex narrative guiding down — has not triggered; if anything AMD's data-center print corroborates it in the opposite direction.

NVIDIA invites a competitive trim and will not get one either. AMD doubling data-center revenue is evidence that merchant alternatives are scaling, a real if modest negative for NVIDIA's share of the accelerator wallet even as it confirms the cycle. But AMD's base is far smaller, and the falsifier written at entry was a capex guide-down or a China export escalation. Neither happened — capex was guided up. Trimming an intact thesis whose relative strength is improving, in order to fund an add into a name that just gapped, is churn dressed as conviction.

What the open has to confirm is narrow: that Tuesday's advance was accumulation rather than a one-session squeeze. Specifically, whether small caps and financials keep participating rather than handing leadership straight back to four mega-caps; whether oil stays lower on the Hormuz headline or reverses when the deal proves less imminent than reported; and whether the AMD pattern — good AI results sold — starts appearing in names this book actually owns. If that pattern spreads, the upgrade was early, and the response is sizing within growth, not a rotation into defensives.

Positioning

Holding	Weight	Thesis state	Action + tree branch	Evidence
MSFT	27%	INTACT	Hold — INTACT branch; adds only on pullback-to-structure, and the name is extended	No company news; Palantir's quarter corroborates the enterprise-AI monetization the thesis depends on
JPM	23%	INTACT	Hold — INTACT branch	No thesis news; momentum60 +10.03% against XLF momentum60 +7.35% confirms sector-wide, not idiosyncratic, trend
AMZN	20%	INTACT	Hold — INTACT branch; decline below the unexplained-decline trim threshold	-2.32% is post-gap digestion ~12% above the 20-day, with no Amazon-specific news in tonight's packet
AVGO	16%	STRENGTHENED	Hold — STRENGTHENED branch, add declined on extension	AMD's data-center revenue doubling is direct evidence hyperscalers are scaling second-source and custom silicon
NVDA	14%	INTACT	Hold — INTACT branch	Falsifier untriggered as capex guided up; momentum60 turned positive at +1.63%, but AMD's share gains cap conviction

No actions — the one state change tonight (Broadcom to STRENGTHENED) resolves to hold rather than add under a risk-on regime that forbids buying leaders after extension. Cash stays at 0% and the book stays fully invested.

What Could Break It

The single headline that would hurt the largest share of this book is a hyperscaler AI capex guide-down. Microsoft, Amazon, Broadcom and NVIDIA together are 77% of the portfolio, and while the IPS sector cap is satisfied at 57% information technology because Amazon classifies as consumer discretionary, the economic exposure is one cycle, not three. That concentration is intentional — it is how the regime call is expressed — but it should not be described as diversified. One guidance sentence from a large cloud buyer next quarter moves three-quarters of the book at once.

The second risk is the one that just improved. The Hormuz headline is a report of a possible deal, not a deal. If it does not land, oil reverses and the supply-side inflation tail returns to the rates driver — and rising yields driven by inflation fear is the single worst configuration for a book this long-duration. That would flip the very input the upgrade rests on.

Third, the AMD reaction pattern. Beating hard and falling anyway is the fingerprint of crowded positioning, and crowded positioning unwinds violently. If Microsoft or Broadcom prints well and trades down the same way, the correct read is that the AI trade's marginal buyer is exhausted, and that becomes a sizing problem across every position except JPMorgan.

Finally, credit is unverified tonight. That is not a risk observation so much as an admission: the framework says an unverified input can only lower risk, never raise it, which is why the regime upgrade came with no increase in gross exposure to the theme.

Institutional Signals

Nothing new was filed — the tracked 13Fs still date to 15 May, and Scion's to 3 November 2025. What changed is relevance, not data. Scion's book was 66% Palantir and 13.5% NVIDIA at that filing, and Palantir is the name that printed the results the tape called otherworldly on Tuesday. That does not mean the position is still held; the filing is nine months old and unknowable from here. But it does say something about where a famously skeptical manager was willing to concentrate ahead of the enterprise-AI monetization proof that is now arriving.

The rest reads as it did. Ackman is concentrated in Amazon at 17.4% and Microsoft at 15.3% — this book's two largest non-financial positions, held by a manager who does not diversify. Bridgewater is spread across index exposure with Amazon at 4.1% and NVIDIA at 3.7%, which is breadth rather than a view. Buffett remains anchored in Apple, American Express and Bank of America at 9.5%, the slow-moving case for money-center exposure that JPMorgan expresses here. None of it is a trade instruction. Collectively it describes managers who never left risk, which is consistent with tonight's regime call and is not, by itself, evidence for it.