Actionable insights — Portfolio Update: Did Last Week Bring Two Seismic Events?

1. Map the resistance above a breakout before deciding whether it matters

1. When a rate or price clears a multi-year level, do not stop at "it broke out." Go find the next levels up and date each one.
2. Rank them by how much history sits behind them: a level "that runs all the way back to" two decades ago is a wall; a level set in a single episode is "a bit of friction."
3. Note the distance between the walls. A wide gap above the nearest one is the real risk — it means there is nothing structural to stop the move for a long way.
4. Attach the macro conditions that produced the highest historical level, so you know what regime would be required to revisit it. If today's conditions rhyme, the far level stops being theoretical.
5. State the honest alternative out loud ("yields may soon peak and even begin to ease") so the map is a contingency plan, not a forecast.

• A decisive close above ~5.5% on the 30-year with no structural level until 5.75% and then a two-point air pocket — and, on the other side, the first failed retest that would define the peak.

2. Separate the trigger from the precondition when arguing a market top

1. Take the historical analogue you want to invoke and decompose it into two independent parts: the vulnerability (what made a waterfall decline possible) and the catalyst (what started it).
2. Assign the causal weight explicitly. A catalyst without a vulnerability is a wobble; a vulnerability without a catalyst can persist for years.
3. Test the two separately against today. Is the vulnerability present? Is a candidate catalyst forming?
4. Because the catalyst is unpredictable and the vulnerability is measurable, position on the vulnerability and use the catalyst only as an alert condition — which is why the instruction is "be on alert," not "sell everything."
5. Watch for the market's own tell: a market "totally unperturbed" by the developing catalyst is evidence the vulnerability is unpriced, not evidence it is absent.

• The first session in which equities and long yields start moving together in the wrong direction — the moment the market re-prices the catalyst it has been ignoring.

3. Swap P/E for Price/Sales when the "E" is the thing in question

1. Default to Price/Sales over Price/Earnings when you need a cross-cycle valuation comparison, for two stated reasons: sales are far less volatile than earnings (so the ratio isn't distorted by where you are in the cycle), and revenue is much harder to manufacture with accounting than profit.
2. Compare the current reading to the most extreme prior episode on the same measure, not to a ten-year average — the question is whether this is the biggest, not whether it is above normal.
3. Expect the two metrics to disagree at exactly the moments that matter: peak-cycle margins make the P/E look reasonable while the P/S screams. The disagreement is the signal.
4. Apply it to single names as well as the index — a shortage-driven, low-P/E, high-P/S name is a peak-earnings trap (the Micron template from Jul-26); a depressed-margin turnaround is the mirror image (the Estée Lauder template).

• Any index or name where the P/E is defensible but the P/S is at an all-time high — and, as the falsification, a P/S that de-rates while sales keep growing.

4. Audit earnings quality by hunting for the customer-equity loop

1. When reported profits are the bull case, ask what non-cash items are inside them before accepting the number.
2. Look specifically for the loop: the seller takes equity or warrants in the buyer as part of the transaction, books the sale as revenue, then marks up the stake and books the gain as income. One transaction is counted twice.
3. Check whether the counterparties are loss-making and dependent on the seller — that is what turns aggressive accounting into circularity, because the customer's ability to pay is itself funded by the seller's investment.
4. Grade it honestly: "legal but… extremely unsustainable." The question is not whether it is permitted but whether it repeats without ever-larger stakes.
5. Discount the reported earnings accordingly, then re-run your valuation on the cleanest available line — revenue.

• The first quarter in which the marks go the other way — stake write-downs flowing through income — and the depreciation on the capex wave finally hitting the P&L.

5. Re-underwrite a margin story the moment the business model changes capital intensity

1. When "margins have been rising for years" is offered as a reason to pay up, ask where the margin came from — here, network effects plus capital-light models.
2. Then ask whether that source is still operating. A company that must now build physical plant is a different business with the same ticker.
3. Quantify the shift with the capex run-rate and its trajectory, not adjectives — and note the growth rate of the spend, which is what forces the returns question.
4. Ask the return-on-invested-capital question explicitly: the spending is only accretive if the revenue it enables exceeds its depreciation and cost of capital. "Serious questions" about that are a valuation input, not a footnote.
5. Remember which way this cuts on the multiple: capital-light businesses earn premium multiples because they are capital-light. Remove that and the fair multiple falls even if earnings hold.

• Capex guidance rising while free cash flow falls, and the first hyperscaler to cut AI capex — the Jul-26 hypothesis that such a cut might send that stock "straight up."

6. Convert a standing forecast into a dated verdict — and say which it is

1. For any multi-year call, restate the original mechanism in one sentence (issuance supply, buyer withdrawal, deficit scale) before claiming the outcome.
2. Then mark it: is this still a prediction, or has the condition been observed? "No longer a prediction but a reality" is a scoring event and should be dated.
3. Do not let the scoring event become the position. A call that has come true is a call whose remaining upside must be re-derived — which is exactly why the next paragraph asks an open question rather than declaring victory.
4. Keep the loser column visible alongside it (the yen call: "it's about time!"). A scorecard that only records wins stops being an instrument.

• The point at which a vindicated macro call has fully paid — long yields at levels where the risk/reward on duration finally inverts and the standing "avoid" becomes stale.

7. Read an official intervention as a positioning event, and grade it by who joins

1. Establish the fundamental contradiction first: an asset at a multi-decade extreme while its underlying fundamentals point the other way (negligible external debt, a current-account surplus, the world's largest creditor position). That is what makes the extreme a positioning phenomenon rather than a verdict.
2. Name the actual mechanism keeping it there — here a central bank "pathetically behind the curve" on rates, which sustains the rate differential the carry trade feeds on.
3. Grade intervention by participation, not size. Repeated unilateral intervention that fails is noise ("an exercise in futility"); the same action becomes a regime change when a second sovereign joins. "What's different this time" is the entire analysis.
4. Check whether you pre-registered the possibility. A move you anticipated in writing is a thesis confirmation; a move you rationalize afterwards is not.
5. Convert to an action with an explicit chase rule: stay long the view, but "we wouldn't chase this rally" — the position is a hedge sized before the event, not a momentum trade entered after it.

• Whether the coordination persists past the announcement, and the escalation trigger Hay names: the BOJ signaling a much more aggressive tightening stance, which converts a managed rally into a disorderly one.

8. Hunt the second-order transmission channel, not the first-order price move

1. After any large move in a funding asset (a cheap-carry currency, a repo rate, a collateral class), ask: what was financed with it? The price move is small; the balance sheet built on top of it is not.
2. Estimate the scale of the dependent position — here, borrowing in a depreciating currency "has been… financing trillions of overseas investments." That scale is why a 4% currency move ranks as a "seismic event."
3. Identify the reflexive accelerant: policy that pulls capital home (repatriation incentives) on top of a funding-cost shock compounds the unwind.
4. Ground it in precedent rather than theory — a prior unwind that produced observable damage proves the channel exists.
5. Rate the crowding: "a macro risk that appears to be off almost everyone's radar" is the condition under which a known channel is still tradeable.
6. Stack the channels. Two independent shocks hitting the same asset ("the one-two punch") is a different risk than either alone.

• Yen strength coinciding with weakness in the most carry-funded assets (the AI complex, EM credit) — the Jul-26 warning that a yen rally would be "a real nasty hit to the AI trade."

9. Run the rating book against the price, not with it

1. Treat the rating as a function of thesis integrity ÷ price. If the thesis is intact and the price has fallen, the correct move is to raise the rating, not lower it — the reverse of what performance-chasing produces.
2. Apply the mirror on the way up: names that have run hard against an unchanged thesis get downgraded to Hold or trimmed, which is how the cash to buy the losers is generated.
3. Make the two sides of the trade explicit on the same page, on the same day, so the book is self-funding rather than directional.
4. When you own both a single name and a basket in the same theme and want less risk, keep the basket and downgrade the single name — the exposure survives, the idiosyncratic risk doesn't.
5. Publish the changes as marked cells with dates, and correct the record when a date is wrong — the scorecard only works if the fill dates are right.

• A Strong Buy that keeps falling with deteriorating fundamentals — the case where "raise the rating into weakness" becomes averaging into a broken thesis; and whether the cash raised from the winners is actually redeployed or held.