Wednesday plan: no trades into the CPI print, with the MSFT-to-AVGO rotation armed

Grade: right process, mildly costly session. Last night's plan was zero actions under R4 with the MSFT trim named but not executed — no thesis broke, so no trade was owed, but carrying 77% long-duration growth into an inflation print cost the book 30 basis points against SPY (-0.62% versus -0.32%). That is the bill for the posture, not a mistake in it.

The Call

Nothing changes tonight. All five positions carry at identical weights into Wednesday — MSFT 23%, JPM 23%, AMZN 20%, NVDA 18%, AVGO 16% — because the CPI report that decides the next month for 77% of this book prints before the open, and every discretionary move I would consider is a bet on that number rather than on a thesis. The MSFT-to-AVGO rotation stays armed, not executed. Second straight zero-action night, deliberately.

Since Last Session

Tuesday's real story was not the index, it was where the selling went. Monday's hard-asset and defensive bid faded: GLD fell 0.39% and XLV fell 0.26%, both giving back part of the prior session's flight. Energy kept leading for a third session, XLE +1.25%, after Iran said the Strait of Hormuz will not open until its conditions are met and US oil moved above $83. Long bonds stopped falling, with TLT +0.16%.

Underneath a modest index decline — SPY -0.32%, QQQ -0.34% — the growth complex as a factor actually held up: XLK was only -0.12%, and IWM at +0.34% outperformed both large-cap gauges. XLF was flat at -0.02%. The damage was concentrated in individual large names rather than in a sector or a factor, with BRK-B -2.46%, LLY -1.37% and COST -0.88% all falling on the same session. VIX closed at 15.28.

The book returned -0.62% against SPY's -0.32%. JPM was the only positive contributor at +0.63%; NVDA was effectively flat at -0.02%; MSFT gave back 0.44%; AVGO fell 1.5%; AMZN was the worst at -2.09% and, at 20% of the book, the heaviest drag on the day.

The one piece of company evidence that mattered arrived after hours: CoreWeave's revenue doubled on accelerating AI infrastructure demand and the stock rose 14%. Riot Platforms signed a deal with Anthropic as bitcoin miners pivot to AI infrastructure. Against that, CNBC reported that Jensen Huang's $500 billion AI financing plan faces a significant risk from China.

Regime Check

• Volatility state: VIX 15.28, the normal band, and lower than the prior session's 15.46. Level unremarkable; direction now mildly favorable for the first time this week.
• Rates impulse and driver: TLT +0.16% on the session, but momentum20 at -1.20% and momentum60 at -2.88% mean yields are still higher over a month. The driver is inflation and supply, not growth — Iran tied any Hormuz reopening to conditions and US oil went above $83. That is the framework's explicitly NEGATIVE row for long-duration growth, and it lands on 77% of this book.
• Leadership breadth: mixed, marginally better than Monday. IWM +0.34% beat SPY -0.32% and QQQ -0.34%, and XLK -0.12% beat the index, so this was not leaders being sold to fund laggards. It was single-name damage inside a flat-to-soft tape. One session is not a broadening trend, but it is no longer clean distribution either.
• Credit and dollar: unverified in tonight's packet. Under the framework an unverified input can only lower risk appetite, never raise it.
• Event proximity: maximum. The CPI report prints Wednesday morning, inside the session this journal covers. Turnover should approach zero.

Regime: R4 — event-suspended. Second consecutive session, so this is a carry-forward, not a regime change, and no two-session confirmation is owed.

Two strongest supporting inputs: event proximity, and the inflation-supply driver behind the rates impulse.

Contradicting input: three of the five inputs improved at the margin — VIX fell, TLT rose, and the defensive bid faded while small caps outperformed — and CoreWeave's doubled revenue is filed-results evidence that AI infrastructure demand is accelerating rather than stalling. If Wednesday's print is tame, the honest read will be that this book was cautious into a non-event, and I will say so.

Plan for Wednesday

The intent is to place no trades before the print. That is the plan, not an absence of one.

The branches after the number are written down now, so the decision is not made in the reaction:

• If core CPI runs hot, the trim of MSFT toward the bottom of its conviction band gets executed and the proceeds go to AVGO. The reasoning is specific rather than directional: at 503.81 against a 20-day average of 435.53, with momentum20 +15.68% and momentum60 +22.33%, MSFT is the book's most multiple-sensitive asset, and a hot print is a multiple event rather than an earnings event. AVGO's contracted hyperscaler ASIC backlog is the least discretionary revenue stream in the AI complex and therefore the right destination for capital leaving a high-multiple name.
• If the print is contained, there is no reactive add. A tame number would validate the existing 77% growth weighting; buying more of what already worked, on the day it works, is chasing.
• Energy is being declined explicitly. XLE has led three sessions and carries momentum60 of +6.70%, and the Hormuz situation is genuinely unresolved. It stays out of the book anyway: it is commodity beta, not a growth leader, and initiating a fresh position hours before a CPI print is precisely the behavior R4 exists to prevent.

What the open must confirm: first, that the AI complex trades on the CoreWeave datapoint and not only on the rates reaction — if accelerating, paid-for infrastructure demand cannot lift the group at all, that tells me the complex has become a pure rates instrument and sizing has to change. Second, that AMZN's decline finds an explanation. I could not source one tonight, and an unexplained gap that repeats stops being noise.

Positioning

Holding	Weight	Thesis state	Action + tree branch	Evidence
MSFT	23%	INTACT	Hold — INTACT → hold, no add after extension	No Microsoft-specific item in tonight's packet; -0.44% versus XLK -0.12% is price, not evidence.
JPM	23%	INTACT	Hold — INTACT → hold	+0.63% against XLF -0.02%, and nothing in the packet touches net interest income or credit quality.
AMZN	20%	INTACT	Hold — INTACT → hold, unexplained-decline flag logged	-2.09% with no Amazon-specific news, but the move sits inside its ordinary daily range and BRK-B, LLY and COST fell alongside it.
AVGO	16%	INTACT	Hold — INTACT → hold, add deferred under R4	No Broadcom news; CoreWeave's result is supportive for the supply chain but says nothing specific about custom-ASIC programs.
NVDA	18%	STRENGTHENED	Hold, add declined — STRENGTHENED → hold or add, add deferred under R4	CoreWeave's revenue doubled on accelerating AI infrastructure demand; offset by reporting that the $500bn financing plan carries China risk.

One state change (NVDA to STRENGTHENED), zero actions. Sector concentration is 57% Information Technology against a 60% cap; the largest single theme is 23% against a 40% cap; era drawdown is -0.03% and the protocol stage is normal.

Institutional Signals

No new filings since 2026-05-15, so nothing here is tonight's evidence and the read is unchanged: Ackman concentrated in MSFT at 15.3% and AMZN at 17.4%, Bridgewater diversified with AMZN at 4.1% and NVDA at 3.7%, Berkshire in financials via BAC at 9.5%. Scion's NVDA at 13.5% is dated 2025-11-03 and over nine months stale — old enough that it should not be cited as agreement with anything.

The one thing worth naming given the tape: Berkshire's CVX at 6.6% is the only energy exposure across the tracked filers, and it long predates the Hormuz situation. A structural position held for years is not a validation of chasing XLE into a third up-session this week.

What Could Break It

The one that matters: a hot core CPI print Wednesday morning. MSFT, AMZN, NVDA and AVGO are 77% of this book between them, and every one of them is a long-duration asset whose multiple is set by the discount rate. A print that confirms the inflation-supply impulse is broadening beyond oil hits three quarters of the portfolio simultaneously, and JPM's 23% is the only offset. No other single headline comes close.

Beyond that:

• A China export escalation on advanced accelerators. This is NVDA's written falsifier, and tonight's reporting on the financing plan's China exposure means the same event would now hit both the demand runway and the funding channel behind it.
• A hyperscaler capex guide-down. It would break NVDA and AVGO on the same day and put AMZN's AWS capacity thesis in question — 54% of the book on one guidance sentence.
• AMZN's decline recurring without explanation. One 2.09% day inside a normal range is noise. A second, with still nothing in the packet, is information asymmetry and triggers the half-trim rule regardless of the regime.
• Credit turning. HY spreads are unverified tonight. If they widen while equities hold, the equity tape is wrong and gross exposure has to come down before the two reconcile — that would also be the fastest route to breaking JPM's thesis, which is the book's only non-AI earnings stream.