Actionable insights — The Big Short Partners Reunite

4:10 1. Stand in the policymaker's shoes and strike out the options he can't use

1. State the actor's problem in his own terms first: "articulate the problem from his perspective" — "40 trillion times an interest rate is a big number… my primary agenda… is to get rates lower."
2. List every real lever in order of effectiveness: end the war (oil "86 to 60"), cut entitlements, raise taxes.
3. Strike each one the politics won't allow: "the president just said, 'For some reason, I can't'"; cut entitlements — "No"; raise taxes — "No."
4. What remains is what he will do, however cosmetic: "then you start with theatricality and deception by playing twist games."
5. Put a clock on it (he needs "six months to nine months" for inflation expectations to fall), then ask what happens if the clock runs out: printing, whether things break or not.
6. Own the asset that wins in both branches: "if something really goes bad… helicopter print money. And if status quo happens… continue printing money."

• A lever reopening (a ceasefire bringing oil down, an entitlement or tax deal); buybacks being upsized; the 10-year's reaction to each signal; any attempt to pull forward lower rates "by signaling."

13:27 2. Read the receivables note, not the revenue line

1. When revenue growth is spectacular, open the 10-Q and find the customer-concentration disclosure — "stuff that we used to do to financial stocks."
2. Look at receivables, not only revenue: who owes the company money at quarter-end. "The top five direct customers of Nvidia accounted for 70% of accounts receivable."
3. Credit-check those customers separately: are they profitable, how are they funded? "A couple of those handful of names are not looking so good. Open AI certainly."
4. Treat the vendor's growth as only as good as its weakest large debtor.

• Receivables growing faster than revenue; days-sales-outstanding stretching; vendor financing, guarantees or equity stakes in the same customers; the top-customer share changing quarter to quarter.

14:38 3. Channel-check the buyer's cost line, not the usage line

1. Ask an operator at a real customer (off the record) two separate numbers: volume and spend. "Queries are up… 30, 40%… But his costs are down something like 60%."
2. When volume and spend diverge, find the mechanism: here, centralised routing that sends only the "very very important queries" to frontier labs and the rest to "the Chinese open-weight models."
3. Scale it with care ("you can't make a whole mosaic on it"), then ask the investor question: if most enterprises do this, "how do we get to the returns on invested capital?"
4. Map the answer to history: a capex boom where usage grows while pricing collapses is the setup for "boom, bust, and then good cycles afterwards."

• Enterprise AI budgets flat or down while usage rises; model-routing and "AI gateway" adoption; frontier-lab price cuts; open-weight model share in enterprise workloads.

17:29 4. Count the new stock coming — supply ends booms

1. Watch the direction of share count across the market leaders: from buybacks ("buying back their own stocks") to issuance ("Google issues $85 billion in equity").
2. Add up the scheduled mega-IPOs and insider lock-up expiries — "they're going to let Anthropic come public because they need to… just like SpaceX."
3. Set it against the demand side, which is mechanical: passive 401(k) flows that only "buy" on money in and "sell" on money out.
4. Treat a rising supply of stock against flat mechanical demand as a late-cycle warning: "that's what usually has killed markets… in 2000, in 1929."

• IPO calendar dollar volume; net equity issuance turning positive across the index; lock-up expiry dates; secondary offerings by insiders of recent listings; 401(k) contribution growth slowing.

23:09 5. For a loss-maker, compare dollar changes in revenue and cost, not growth rates

1. Put two competitors' sequential growth side by side: Anthropic ~$11.5B, "up over 100% in 3 months," vs OpenAI $6.5B, "up only 18%."
2. Then drop the percentages: "forget about… the percentage increases, if you just look at the dollar changes in three months their revenue went up a billion and their cost went up three."
3. If incremental cost dollars exceed incremental revenue dollars, the business is scaling its losses, whatever the headline growth.
4. Add the funding test: an unprofitable firm needs its story intact — "narrative is everything" — so a turn in narrative raises its cost of capital just when it "can't afford that."
5. Look ahead to the resolution history suggests: the asset survives, the equity doesn't ("whoever buys OpenAI out of bankruptcy… a fantastic deal").

• Incremental cost vs incremental revenue each quarter; share-loss data; senior departures; down-round or structured-funding terms; S-1 disclosures once filed.

30:20 6. Before shorting, know who else is short and how long they can hold

1. Recognise the dominant short seller today: multi-manager "pods" running ~5:1 leverage, beta-adjusted and factor-neutral, which must keep volatility near 6–9%.
2. Infer their horizon from their constraints: they bank a 3–5% spread and move on, and "if it doesn't work for them in 3 weeks, they're covering." A levered book "can't" lose 4–5% a month.
3. Assume any obvious fundamental short is already crowded by them "in size," so squeezes come on their exits, not on your news.
4. Remember what a factor-neutral book is protected from: concentrated factor bets (long AI / short software) get taken out on one reversal — Eisman's 1905 car-crash analogy.
5. Size your own short for that path (the partners now short "with significantly less capital") or skip it. A "cult" stock (Tesla) fails this test by default.

• Short interest and borrow cost rising on your name; prime-broker reports of multi-manager crowding and gross leverage; sharp factor reversals (momentum, growth/value) that force pods to cover.

37:33 7. Back out the price an unnamed buyer must be paying

1. Find what share of profit comes from selling assets rather than holding them: gain on sale is "between 75 and 100%" of Carvana's pre-tax income.
2. Get the one disclosed market price: the known whole-loan buyer (Ally) pays "between 102 and 104."
3. Compare it with the reported average (109–110) and solve for the rest: if ~20% goes at 102–103, "somebody in the world is paying north of 110… 112, 113."
4. Look for who that could be among related parties: large holders, their insurers and vehicles ("the number five holder… Delaware Life"), and court or regulatory filings showing related-party exposure ("from like three to something like 30 or 40%").
5. Name the catalyst that would expose it, since short sellers are paid by disclosure: "if Delaware Life goes under… the buyer of the paper is gone."

• Gain-on-sale margin vs the disclosed buyer price each quarter; changes in the buyer list; insurance regulator filings on related-party assets; any stress at the suspected buyer.

40:17 8. Only expect to get paid on a short when something forces disclosure

1. Separate being right from being paid: a short pays "if you bring light… or truth to what happens."
2. For each short, name the event that makes the hidden fact undeniable: losses that can't be buried, marks that can't be faked (the Big Short moment), a funder failing, a regulator acting.
3. If no such event is in view, expect "a lot of nothing" and squeezes. Size down or wait.
4. Favour shorts where a public actor is already pushing the disclosure: a regulator attacking pricing (Pulte on FICO).

• Regulatory orders, lawsuits and court filings; auditor or rating-agency changes; a funder's distress; the first quarter a hidden cost shows up in reported numbers.

42:30 9. Own the low-capex toll road into a country turnaround

1. Start with a country where policy has changed and resources are trapped: Milei "figured out their fiscal situation and balanced the budget," and the gas is there, but "how do you transport it out?"
2. Find the bottleneck and compare how to fix it: a $30–50B onshore plant (YPF, the Cheniere model) vs a floating unit at "three and a half or two and a half billion."
3. Prefer the asset that arrives first, costs least, and earns a contracted fee: "basically a 20-year contract, a toll road."
4. Link it to the macro loop: exports → budget surplus → falling rates ("100 plus% to 30, come down a lot more").

• Vessel delivery and first-cargo dates; contract counterparty credit; Argentine election and policy continuity; Argentine rates and the budget balance.

45:40 10. Buy the low-cost producer before a rule change opens a price gap

1. Find a product whose price differs hugely between markets because of regulation: "$200 a pound" in California vs "600, 700 bucks a pound" in Germany.
2. Own the lowest-cost producer in the cheap market ("90 to 100 bucks a pound"), which profits most if the gap can be crossed.
3. Track the rule change step by step (Schedule I → III done; interstate or export sales "not yet") and write down margins if it happens: "50 to 65% EBITDA margins."
4. Be honest about timing ("a 2027 event") and look for small real signals first ("they sold hemp overseas this quarter").

• Export licences, EU medical-cannabis import approvals, interstate commerce rules; the overseas share of sales; California wholesale prices.

47:21 11. Touch household names only when they are "for sale"

1. Spend most of your time in small caps "that no one else really traffics in."
2. Go into mainstream names only when a sector-wide narrative is dumping them: "the only time we go into… the mainstream… is when something's for sale."
3. Check that the narrative is about disruption fear rather than current numbers ("when Google was allegedly being disrupted"; the SaaSpocalypse).

• Sector-wide drawdowns driven by one narrative; whether the reported numbers actually break; the fear fading.