# Notes

> Source: https://sentity.co/notes.html
> Author: Jason (@JasonSentity)
> Notes are dated and are not edited after publication. A wrong thesis stays up.

## Warsh Did It

*September 23, 2026* · https://sentity.co/notes/2026-09-23.html

Warsh did it. On 16 September the FOMC raised the funds rate 25 basis points to 3.75–4%, the first hike in more than three years, on a unanimous 12–0 vote. His line at the press conference: inflation has been **"too high for too long."** Sixteen of eighteen participants expect another increase this year; four think two. Warsh himself declines to submit a dot.

So the tug of war I wrote about in August resolved the way the hawkish read said it would. He was not bluffing.

The bond market did the rest. The ten-year went through **5%** on 14 September for the first time since 2023, and has kept going — 5.10% this week on a 13 basis point day, the largest single-day move in about eighteen months and the highest yield since 2007. The drivers are exactly the three I named earlier this month: inflation with a war premium on energy, corporations issuing debt to fund AI buildouts, and a government issuing to cover everything else. Nothing exotic. Just three large borrowers and one pool of savings.

Which puts us in the part of the cycle I have been waiting on. Rates this high, held this long, eventually find whatever was built assuming they would stay low. So I have been scanning the horizon for cracks — and this is the kind of work I now hand to AI first, because it will read a hundred filings, delinquency series and rating agency notes in the time it takes me to read three.

**It found one, and it is not the one I expected.** Not commercial real estate, though that is bad and getting worse — CMBS office delinquencies have gone from 6.5% at the start of 2024 to **11%** today, and mid-sized banks carry CRE at roughly 190% of their tier-one capital, with the worst quintile above 350%. Everyone knows about CRE. It has been telegraphed for two years.

The one that got my attention is **private credit**. The default rate hit **6% in April** by Fitch's count, in a market that has grown to roughly $2 trillion with almost no public price discovery. And the stress is concentrated in corporate direct lending — specifically, in **software**.

Sit with that for a second. The most levered corner of the credit market is concentrated in the one sector AI is actively disrupting. These are companies that borrowed at low rates against recurring revenue everyone assumed was permanent, and are now refinancing at 5%-plus into a market where their product may be a feature in someone else's model. That is not a rate problem alone. It is a rate problem arriving at the same moment as a business model problem.

The contagion path is not a bank run. It is slower: defaults accelerate in direct lending, the losses surface in regional banks, insurers and pension funds that bought the paper for yield, and those institutions pull back everywhere at once. Nothing breaks loudly. Credit just quietly stops being available to anyone who is not investment grade.

I have no position here and I am not shorting anything. But it changes what I am watching. If private credit is where this cracks, the tell will show up in regional bank disclosures and BDC marks well before it shows up in the S&P. That is now on my weekly list.

Positioning unchanged: long the AI names, large cash, gold. If this is the crack, cash is what lets me buy the good companies that get sold alongside the bad ones.

---

## Higher Rates and Inflation Until Something Cracks

*September 10, 2026* · https://sentity.co/notes/2026-09-10.html

Here is my base case, stated plainly so I can be held to it later.

Inflation is not finished. Iran is keeping a bid under energy, and energy leaks into everything — freight, food, utilities, the price of running a factory. On top of that, the fiscal side keeps creating money. The deficit runs without any serious attempt to close it, and now there is a proposal to hand every American $5,000 — roughly $1.6 trillion, every dollar of it borrowed. It is hard to call that anything but reckless. You do not mail a check to every household in an economy that still has an inflation problem, and you certainly do not borrow $1.6 trillion to do it. That is demand poured straight onto the fire, and the bill lands on the same bond market that already has to absorb everything else.

Meanwhile the Fed is going the other way. Warsh has been about as clear as a Chair gets that he intends to hold the line. So you get a tug of war: Treasury and Congress pushing money into the system, the central bank trying to pull it back out.

I don't think that resolves cleanly. I think rates creep higher. Not a spike — a grind, quarter by quarter, driven by the two borrowers I keep coming back to: AI capex that has moved onto the bond market, and a government that will not stop issuing. Both are price-insensitive in the near term. Both crowd out everyone else.

And that is how this ends. Rates climb until something in the economy cracks — housing, commercial real estate, small business, the leveraged borrower who has been rolling debt and praying for relief. Not the AI names. They can pay these rates and will keep spending straight through a downturn, because for them falling behind is the more expensive outcome. It is the rest of the economy that breaks first.

So the positioning follows directly. Stay long the AI names; that is the part of the economy that compounds through the downturn rather than in spite of it. And keep a large cash position — not as a forecast, but as an option. If the crack comes and good names get beaten up alongside the bad ones, I want to be buying rather than watching. The cash does cost me something every month it sits there, but that is the price of being able to act.

---

## AI Can Pay the Rate. Can the Rest?

*September 1, 2026* · https://sentity.co/notes/2026-09-01.html

AI has discovered the bond market. What used to come out of operating cash flow — data centers, power, chips — now increasingly arrives as issued debt, and it comes in size. That paper competes for the same pool of savings as everything else, and the largest borrower in that pool is the U.S. Treasury.

Treasury's need is enormous and it is not going away. The deficit runs, the maturity wall keeps rolling, and every dollar of it has to be placed with somebody. The somebody on the other side is no longer a captive buyer. With inflation still near 3.7% and no clear path down, lenders are asking a real price for tying up money for ten or thirty years. Add a new borrower with a genuine story and enormous appetite, and the price the government pays goes up too. That is crowding out, and it doesn't need a recession to hurt.

Two stories are competing from here, and I think both deserve weight.

**AI is real.** The spending is not a discretionary line item that gets cut in a soft quarter. For the companies doing it, falling behind is the greater risk, so the capex continues through economic downdrafts — funded by debt if cash flow won't cover it. If this is right, a narrow slab of the economy keeps compounding no matter what the rest of it does.

**The rest of the economy may finally stumble.** Rates this high, held this long, eventually reach everything that has to refinance — housing, small business, commercial real estate, the marginal borrower everywhere. That process has been slower than anyone expected, but slower is not the same as avoided. High rates plus a bigger crowd competing for the same capital is how it would arrive.

The two are not mutually exclusive. The most likely world is a boom in one narrow place and a grind everywhere else — which is roughly what the market is already showing, only more so.

So: larger cash reserves than feels comfortable, because that is what turns a stumble into an opportunity. Keep gold. Keep tech and AI — that is where the productivity actually is. Space is more speculative and I hold it with clear eyes, but I'm not selling Rocket Lab before Neutron flies; that is the whole thesis and it hasn't been tested yet. And if things get dicey, the plan is not to hide. It is to accumulate the great AI names while everyone else is selling them.

---

## Warsh vs. Bessent

*August 29, 2026* · https://sentity.co/notes/2026-08-29.html

Warsh gave his first Jackson Hole speech as Chair on Friday, and it was about as hawkish as a Fed Chair gets without pre-committing to anything. He called the 2 percent PCE objective **"a firm, fixed target."** On the better summer prints, he refused the obvious dovish read:

> "And while this summer's PCE and CPI readings were better than expected, they do not tell me that underlying trends have meaningfully improved."

Then the sentence that matters: **"We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do."** With PCE near 3.7%, the market heard it correctly and moved to price a September hike as more likely than not.

Treasury is pulling the other way. Bessent has doubled long-end buybacks, aimed at the 10- to 30-year sector where there has been a buyers' strike since June, with talk of tapping the Treasury General Account to fund more of it. Call it what you like — it is an effort to push the long end lower.

So the fiscal side is easing while the monetary side is tightening. Two hands on the wheel, turning opposite ways. That is not a stable arrangement, and it resolves one of two ways.

**Warsh follows through.** **"I stand here today committed to a discipline, not to a decision"** is the language of someone who intends to hold the line and does not want to be argued out of it. Rates go up or simply stay restrictive, the buybacks prove too small against the supply that deficits keep producing, and financial conditions tighten. Growth slows from here. But slower is not the same as slow — the AI tailwind is real, and genuine productivity gains can absorb a surprising amount of monetary tightening. This could land as an actual slowdown, or as nothing worse than a cooler version of the boom. That is a wide range, and I don't think anyone knows where it lands.

**Warsh capitulates.** The pressure from the White House is relentless and the long end refuses to cooperate, so policy ends up looser than Friday's language implied. Inflation gets tolerated rather than beaten, and the debt gets inflated away roughly on schedule.

I don't know yet which way this breaks, and that is exactly why cash is a large position. The real assets I already own — stocks, gold, and bitcoin — cover capitulation. Cash covers the other: if growth slows in earnest and prices come down, I want to be able to act. If the AI tailwind wins out instead and everything grinds higher, cash is a drag and I will have paid for insurance I did not need. With stocks priced where they are, I will take that trade.

---