Actionable insights — Weekly SSR: hedging a hike with the asset priced off the policy rate, selling a proxy once its option is paid for, buying a convert-arbitrage flush, and passing on a spread that only works with borrowed money

1. When the rate call flips, hedge with the asset priced off the part of the curve that will move — not the one in the headlines

1. Admit the change in your base case explicitly and date it. A hedge put on while still defending the old view is sized for the wrong world.
2. Write both branches of the decision and who loses in each. If the central bank hikes, which borrowers reprice immediately? If it does not, which assets reprice through the long end instead?
3. For each branch, identify the benchmark the damaged asset is priced off. Residential mortgages follow the 10-year; commercial property loans, leveraged loans and small-cap floating debt follow SOFR, which tracks the policy rate almost one-for-one. A hike can lower the 10-year (growth fear) while raising SOFR.
4. Within the exposed sector, screen for concentration against capital (for banks, non-owner-occupied CRE to total risk-based capital against the 300% supervisory line), low non-interest income to absorb losses, reserves below peers, and dependence on wholesale funding.
5. Check the quality of any de-risking, not just its size: a lender that cut its concentration by selling performing loans has a worse residual book than the ratio suggests.
6. Time the entry to price, and size the squeeze risk before shorting: prefer a name near its highs with modest short interest and moderate days-to-cover over a crowded short.

• The FOMC statement on September 16 and whether the path is priced as "one and done" (his view) or two hikes (the desks'); SOFR and the 2-year versus the 10-year — the hedge works on the front end, and the call closed with the 10-year falling on growth fear while the Russell was the least-weak index; the bank's next quarterly CRE concentration disclosure (the call and deck print 317-333%); any extension of criticized-loan or provision guidance; and the rating agencies — S&P's positive outlook predates any expectation of a hike.

2. Own a listed wrapper around a private asset only while the private asset is free — and compute the implied private mark to know when it stops being free

1. Find listed companies holding a meaningful stake in a private company that investors cannot otherwise buy — the wrapper is the trade.
2. Value the operating business on its own, neutrally, as if the stake did not exist.
3. Subtract that from the wrapper's market value. Divide the remainder by the stake size to get the private valuation the market is implicitly assigning.
4. Enter when that implied mark is near zero or far below credible private-round marks: the stake is optionality for free.
5. Exit when the implied mark reaches or exceeds the most optimistic credible valuation — at that point you hold a fairly valued core business plus downside to any markdown, not an option.
6. Pull the exit forward when a catalyst for the private asset (a listing, a round) becomes less certain, and state plainly that you can re-enter lower.

• Anthropic's S-1 timing and reported pricing range (the deck says an October 2026 target at ~$2T; the call says October 26th); secondary-market marks on the private stake; the wrapper's own dividend and Korean-market discount, which dominate once the option is priced; and a sell-off that brings the implied mark back well below credible valuations — the re-entry condition, stated in the deck title as "Can Re-enter Later."

3. Classify a scary headline as sentiment or fundamental — and write down in advance the evidence that would reclassify it

1. Read the primary document, not the coverage. List what it actually commits to: numbers, dates, spending, launches.
2. If it changes none of those, it is sentiment risk. Expect multiple pressure and "chop," but do not trade it as an earnings event.
3. Compare the speakers' incentives. A caution that costs one party nothing and another party a great deal (because only one is about to raise money) carries different information from each.
4. Note the defensive tell: when an executive feels obliged to explain that a word does not mean what it sounds like, the market will spend days debating it anyway.
5. Write the reclassification triggers as concrete, observable events — a delayed product, a pushed deployment, lower component demand, a guidance cut citing the theme.
6. Separately, trace the second-order funding chain that could turn sentiment into fundamentals later, and position lightly for that path (sell the most exposed proxy) rather than for the headline.

• The four named triggers, one by one; hyperscaler capex guidance at the next prints (none had moved as of the call); the Anthropic IPO date and pricing; OpenAI financing headlines; semiconductor-versus-energy relative performance on the Monday open, which he flagged as the first positioning read; and the 2027 S&P EPS estimate ($416) — the place a real slowdown would show up first.

4. Buy the forced selling from a small issuer's convertible — size the hedge flow against the market cap before you size your position

1. Screen for convertible issuance by small-cap companies, especially where the convert is large relative to market capitalisation and daily volume.
2. Estimate the hedge flow: convert size × initial delta (roughly 0.5-0.65) = dollars of stock convert-arbitrage buyers must short. Compare it to the market cap and a day's liquidity.
3. If the flow is a large fraction of the float and must be done fast, a sharp drop is mechanical. Confirm what else was announced the same day (asset sales, book-value hits) and whether it is material.
4. Check the purpose and terms: refinancing a nearer maturity at a lower coupon and a premium near book is balance-sheet repair, not distress.
5. Look for an issuer buyback run concurrently — a buyer of last resort into the hedge selling.
6. Apply a management filter (capital-allocation history, alignment with holders) and check the credit quality of the underlying assets, then size for the risk that remains (here, interest-rate marks) and add in steps rather than all at once. Consider the issuer's senior exchange-traded debt as the lower-risk expression.

• The final conversion price and how much of the 2027 bond is actually retired (UBS's open item); execution of the $20M concurrent buyback and the remaining ~$111M authorization; short interest and borrow data normalising as the hedge settles; mortgage rates (the named residual risk, near 6.85-7%); and the next quarter's book value, since the thesis is "price below book, not a credit event."

5. Publish a DCF on a contested buyout — then discount your own number and trade the vote math, not the fair value

1. Build the valuation transparently: cost of equity from CAPM (current 10-year as the risk-free rate, an explicit equity risk premium, the company's beta), cost of debt, free cash flow to the firm, terminal value, minus net debt.
2. Add an explicit small-cap premium to the discount rate rather than an unexplained haircut at the end.
3. Then list the risks the model does not capture (disruption exposure, acquisitions, leverage) and state that the fair price should trade at a discount to your output. The model sets direction and magnitude, not a target.
4. Identify the governance mechanism that can force a better outcome — for a controlling-holder buyout, the majority-of-the-minority vote — and do the arithmetic of how many independent votes the objectors still need.
5. Name the loophole: decisions the controller can take without that vote (acquisitions, capital deployment) erode the thesis and explain part of the discount.
6. Harvest partially into strength, keep the remainder for the bump, and publish the level at which you would rebuild.

• Special-committee communications and any revised bid; additional holders joining the objection (the gap to a blocking minority); further acquisitions announced by the controller; leverage covenants as debt service rises with rates; and — for the model itself — how much of the output is terminal value, since a 2% change in the discount rate on a small cap moves the answer far more than a year of cash flow.

6. Convert every arbitrage spread to an annual return and compare it to your unlevered hurdle — a spread that only works with borrowed money is not your trade

1. Express the spread as an IRR to a realistic closing date, not as a percentage to the deal price. Regulatory reviews stretch dates; use the later estimate.
2. Estimate the break probability and where the target falls if the deal breaks (the unaffected price, adjusted for what has changed since).
3. Note who is setting the spread. When large funds that use leverage are "putting the spread on," the market-clearing IRR is set for levered capital.
4. If your book does not use leverage, require an unlevered return that compensates for the break risk on its own. If the IRR does not clear it, pass, however likely the close.
5. Prefer situations whose width comes from size, complexity or a contested process — small caps, two bidders, a controlling holder — where levered institutional capital does not compress the spread.
6. For those, bracket explicitly: unaffected price (floor), current bid (base), topping-bid range (upside), and check for a financing condition.

• ROKU/FOXA's second-request timeline and any widening that pushes the unlevered IRR past his threshold; the RSVR special committee's response to Irenic and any definitive agreement (there is "no close date… because it's not clear which bidder is going to win yet"); the gap between the RSVR price and $10.50 narrowing, which reduces the remaining reward; and the general rule's failure mode — a small-cap spread wide for a reason (financing, a controller who can walk) rather than for size.

7. Before pricing a takeover, normalise the earnings base the buyer will pay for

1. When a company announces a sale process, pull the holder list first: who controls the vote, and whose exit is this (a sponsor, a group of asset managers, management)?
2. Compute the multiple on both forward and trailing earnings. A large gap between them means recent earnings jumped, and the takeout price depends on which base a buyer accepts.
3. Decompose the jump: volume and contracted fees (durable) versus commodity-price participation or one-offs (cyclical). Buyers pay full multiples only for the durable part.
4. Get outside calibration — analysts covering the sector, precedent transactions — before estimating a bid.
5. Only then compare the implied takeout range to the current price. Until you can, keep it on a watch list and say so.

• The split of recent EBITDA growth between fee-based throughput and commodity-linked margin; Waha and Permian gas prices, since the commodity component moves with them; named bidders (the deck mentions ET, TRGP and PSX, and earlier Western Midstream interest) and whether they are strategic buyers who could pay for synergies; Blackstone's Class C voting control (close to half the vote per the deck), which decides whether the sale happens on the sponsor's timetable; and the follow-up write-up after the analyst calls.

8. Read an index near its high through breadth and multiple decomposition — then make the cheap-looking forward multiple conditional on the growth that supports it

1. When the index barely falls despite a stack of adverse macro news, check participation: the share of members above their 20- and 50-day moving averages, and the equal-weight index versus cap-weight.
2. Identify the ballast — how much of the index weight is doing the holding up, and how far it has rallied.
3. Decompose the year's return into earnings growth and multiple change. A flat-ish index can hide a large de-rating offset by earnings.
4. Strip out one-off earnings (unrealised gains on private stakes) before accepting the growth figure.
5. Compute the forward multiple on next year's estimate, then ask what that multiple becomes under the specific risk you are worried about. A "cheap" forward multiple is only cheap if the forecast survives.
6. Respect the counter-evidence: if earnings have beaten all year, estimates may keep rising for a while before reality tests them.

• Breadth recovering or failing to recover on any index bounce; the equal-weight S&P's broken uptrend (Fundstrat); revisions to 2027 EPS ($416) — the first number to crack if the AI funding chain slows; the Mag 7 multiple versus its 2022 lows; and the oil-yield correlation (0.85), which is the channel through which the energy shock reaches valuations.

9. On a contested macro call, print the strongest quantified opposing framework — and note exactly which new fact it predates

1. State your own number and horizon.
2. Find the best-argued opposing view that comes with a model, not just a conclusion — ideally one built from an observable input and a published elasticity.
3. Reproduce its arithmetic: the input, the sensitivity, the implied price.
4. List its qualitative reasons separately from the math, so you can test each.
5. Identify which facts arrived after it was written, and whether they change its input — that is where your disagreement actually lives.
6. Keep both views in front of subscribers; update toward whichever input proves right.

• Actual Persian Gulf export volumes (Brooks' input) after the pipeline strike and after any repair; Yanbu loadings (2.9-3.7 mb/d before the attack); Hormuz transits (5-10/day vs 85 pre-crisis) and Bab el-Mandeb transits (26-28/day vs ~70); Saudi output (6.2 mb/d, the lowest since 1990); the rescheduled GCC-Iran meeting and any US position on a transit corridor; and tanker day-rates as the market's own real-time read of rerouting.