Actionable insights — Extra newsletter (September)

1. "Cut what?" — rank the budget by what can't move, and the leftover lever is your forecast

1. Take receipts for the fiscal year to date and set against them the lines that are politically or economically fixed: entitlements, veterans' benefits, net interest, then defense.
2. Express them as a share of receipts, and compare their growth rate with receipts' growth rate. If the fixed lines alone approach 100% and grow faster, every other function of government is already borrowed.
3. Check where any claimed "savings" actually came from. Accounting re-estimates and small agencies do not count if every line big enough to matter still grew.
4. For each candidate cut, ask what it does to receipts. With federal spending near a quarter of GDP, a cut big enough to matter causes a recession that widens the deficit.
5. Whatever lever is left is the policy forecast. Here that is rates, and only in one direction, with the currency as the pressure valve.

• Any "credible fiscal path" announcement. Ask which of the three lines it cuts and how many votes it wins before a midterm.
• The dollar, because that is where the pressure vents if rates are held down.

2. Test whether a rate hike still tightens: who receives the interest?

1. Identify the inflation's source. Bank-lending-driven inflation (1970s: boomers borrowing, low public debt) is throttled by hikes. Fiscally driven inflation with public debt above 100% of GDP is not.
2. Follow the interest payments. When government debt dwarfs private credit, a hike raises government interest paid to savers (retirees in money-market funds), so "part of every hike comes back as stimulus."
3. Stress it with the extreme case. Set the policy rate at the historical maximum you are being told to fear, multiply by the debt stock, add the fixed outlays, subtract receipts.
4. If the extreme case produces a deficit bigger than the tightening it was meant to deliver, the tool is broken for this regime. The remaining drivers (energy margins, medical costs) will dominate the inflation path.

• Net interest growth rate against receipts growth (here +12% against +3%).
• Crack spreads 3–6 months out, and employer health-cost surveys (+8.2% for 2027). These are inflation drivers the policy rate cannot reach.

3. When the natural buyer doesn't show up at an attractive price, look for what it can't sell

1. Name the buyer who historically backstops the market at this level. At a 5–5.5% long bond that is pensions and life insurers locking in decades of liability funding.
2. If they are absent at a price they "should" love, the obstacle is on their balance sheet, not their view of the asset. To buy, they must sell something.
3. Find the asset they would have to sell and ask whether selling it forces a mark. Unmarked private placements (~23% of life insurers' admitted bonds, up from ~18% in 2021) only reveal losses when sold.
4. Work out the reflexive failure mode. An insurer facing a capital hole sells its most liquid asset, which is the very bond you expected it to buy.
5. Then ask who can buy without selling: regulators freeing bank balance sheets, and a central bank creating reserves. That is where the eventual bid comes from.

• Private-credit gating and solvency problems "at the margins": the point where the unmarked assets get marked.
• Bank leverage-rule loosening and any growth in the Fed balance sheet. Those buyers don't have to sell anything first.

4. Score a policy intervention against what the market expected, not against zero

1. Record the expectation before the announcement, not just the size. Here dealers braced for $7–10bn per operation.
2. Compare the announcement with that expectation. A doubling that undershoots the expectation is a disappointment, and the long end sells off.
3. Then watch execution against the intervener's own cap. Buying $5.19bn against a $6bn cap with $10.5bn offered was only the third shortfall in 53 operations.
4. Read the price response as a vote on credibility: 11bp higher on the day means the put is being tested, not respected.
5. Ask what early intervention signals. Stepping in with no liquidity seizure risks a Streisand effect: "the harder Bessent stares at the long end, the more investors wonder what he sees that they don't."

• Each operation's take-up against its cap through early November: "every operation the market deems too small chips away a little more at the Treasury put."
• How the buybacks are funded (bills, TGA cash), which shows whether the program is liquidity support or disguised maturity management.

5. Check a growth narrative against the tax base it depends on

1. Split government receipts by source. Here ~70% (~$3.4tn of $4.85tn) came from paycheck withholding, and individual income plus payroll taxes made up ~87%.
2. Ask what the popular growth story does to that source. AI's payoff "rests on corporate America doing more with fewer white-collar workers," which shrinks the payroll base that funds the budget.
3. Add the capital-competition channel: record corporate bond issuance to fund the same capex draws from the pool the Treasury taps.

• Withheld-tax receipts against white-collar employment.
• Hyperscaler bond spreads widening alongside Treasury yields, which would mean they are competing for the same buyers.

6. In a buyer–seller standoff, check the calendar and where the bid keeps appearing

1. When spot volume dries up, separate caution from capitulation. Are sellers softening their offers? If not, the standoff is buyers waiting, not sellers giving up.
2. Check what the buyers are waiting for. Utility fuel buyers wait for new budget cycles (October) and industry events (NEI Houston). A standoff early in the year can last; late in the year, with fresh budgets, it tends to break.
3. Map the price where the bid reliably reappears on dips, and compare it with the support level contacts described. A dip-buying level that rises over time is a floor moving up.
4. Cross-check against the less-watched price. If the long-term price keeps setting records on thin volume, "waiting for replacement rate volumes before expecting price movement gets it backwards."
5. Use sentiment extremes to act against the equity tape, and pre-commit the next add level.

• October budget-cycle contracting and the NEI fuel seminar, the dates he expects to end the standoff.
• The July low in the uranium equities, his pre-announced add level.

7. Bracket a headline tonnage with grade scenarios before you let it into your model

1. When a deposit is announced as tonnes of ore with no grade, compute the contained metal at a low and a high plausible grade.
2. Note what else is missing: recoverability, annual output, whether the metal is a byproduct of another operation (which caps its output at the host mine's pace).
3. If the bracket spans "footnote" to "meaningful," file it as a headline, not supply, until a grade is published.
4. Apply the same rule-of-thumb check to demand announcements: GW × pounds per GW gives annual pounds, plus initial cores.

• A published grade and recovery for any sovereign "discovery" before it enters a supply model.
• Life-extension decisions, which turn into fuel demand immediately.

8. Read a resource update through ounces, grade, tonnes and dilution, not just the headline total

1. Compare the new estimate with the last one on each axis separately: category ounces (Indicated, Inferred), category grades, and total.
2. Ask why the grade moved. A stricter classification (reasonable prospects for eventual economic extraction) ahead of an economic study is responsible, not a red flag.
3. Back out the tonnage (ounces ÷ grade) and set it against the vein widths drilled. Narrow veins mean dilution will cut grade at the mill.
4. Price the market's valuation per in-ground ounce against the metal price to see what the market is assuming about development.
5. Count the undrilled upside separately (known veins not yet at resource level, untested anomalies), and judge the result against the rocks rather than the hype.

• The dilution assumption in the first economic study.
• Drilling on the 12 veins not yet at resource level.

9. When the price deck is stale, compare the study's assumptions with spot before trusting the market's discount

1. Write down the metal prices a study used and the cost per unit it produced.
2. Compare them with today's spot prices. If spot is far above the study deck, the published NPV understates the asset; if the stock is still falling, the market is pricing something other than economics.
3. List what that something is (jurisdiction, capex, time to production, a partner leaving) and judge whether the drilling news changes any of it.
4. Check whether new drilling is grade-accretive in the tonnes the next study will count. Infill holes at double the M&I grade matter most if they land in the early years of the mine plan.

• Whether the high-grade intervals fall in the early-years pit, which changes payback most.
• The PFS price deck. If it uses anything near spot, the study rerates the asset on paper.

10. Size a government concession in dollars today, then read it as a signal about tomorrow

1. Apply the new terms to current qualifying volume only, and compute the annual saving honestly, even if it is small.
2. Then look at the terms on future and unproven barrels. A flat low rate on untested formations is the government pricing in future investment.
3. Note the conditions (wells to be drilled within a set time or the concession reverts). That sets the date you can check.
4. Ask what the concession says about how the state sees the asset. "Governments do not hand stranded assets royalty cuts."
5. Set it against the valuation: cash, debt, margin per barrel and quarterly cash flow against market cap.

• The two required Block 131 wells, and the October drilling restart at Bretana.
• The erosion-control contract signed before the river rises.

11. After a small-cap collapse, value it against cash, name the one de-risking event, and own the sizing error

1. Separate the drivers of the drop: a strategy change, a supply overhang (lockup expiry), and sentiment. Only the first says anything about the business.
2. Compare market cap with net cash, burn and runway, and account for any convertible preferred whose conversion price has reset. What is left is the market's price for the technology.
3. Identify the single event that de-risks the story and what the last comparable event did to the stock (here CDR, a 25–30% one-day pop).
4. Work out what the upside case implies in company value (50–100x from $2 is $10–20bn) and name what would justify it (fleet production contracts plus civil certification).
5. Be honest about the error. Sizing it as a small, very high-risk position was the defence; underestimating how violently a post-SPAC pre-revenue name can fall was the mistake.

• A date for the C-130J flight demonstrations, and any production orders.
• Further preferred conversion resets, and burn against the 2028 runway.