In late March, the man who runs the Federal Reserve's trading desk stood in front of the Money Marketeers of New York University to explain what his desk had been buying. Roberto Perli, manager of the System Open Market Account, confirmed the numbers that had been accumulating quietly in the Fed's weekly statements: since December, the desk had been purchasing $40 billion of Treasury bills a month. The program has a name, "reserve management purchases," and Perli's speech is largely devoted to explaining why that name, and not the other one, is the right one.
This article does three things: lays out exactly what the Fed has done since October, weighs honestly whether it deserves to be called QE, and checks the result against the prediction this website logged in June, which said the next big accommodation would arrive dressed as a technical fix. The short version of the check: the dress is confirmed, the size is not. Gradual, so far.
One caution up front. The June article, Gradual, Then Big, and the scorecard entry that came from it are the author's own claims, so this is a self-graded progress report, not a resolution. The mitigation is that every figure below comes from a primary source (two FOMC statements, one implementation directive, one New York Fed speech, and the Fed's own weekly H.4.1 balance-sheet release) and the deltas are computed from the same series, so anyone can rerun them.
What the Fed Did
The sequence ran from late October to mid-December. On October 29, with bank reserves, the cash commercial banks keep parked in their own accounts at the Fed, at $2,848 billion (their lowest level since January 2023) and the Standing Repo Facility, the Fed's emergency cash window for dealers, lending $29.4 billion in a single day, the Federal Open Market Committee cut rates a quarter point and announced it would "conclude the reduction of its aggregate securities holdings on December 1." Quantitative tightening, the three-year program of letting the balance sheet shrink, was over.
On December 10 the committee cut again, to 3.50-3.75 percent, over three dissents, and added a sentence that mattered more than the cut: the Fed "will initiate purchases of shorter-term Treasury securities as needed to maintain an ample supply of reserves." The directive took effect the next day: Treasury bills, and if needed anything with three years or less remaining. The desk started buying at $40 billion a month, deliberately front-loaded, Perli explained, to get ahead of the April tax season, when payments to the Treasury drain reserves out of the banking system.
The result, from the weekly H.4.1 statements: the balance sheet bottomed at $6,535.8 billion on December 3 and stood at $6,724.6 billion on July 1. That is $189 billion of growth in seven months, and the direction has one plausible way to continue: Perli's own speech notes that the Treasury's cash account (the government's checking account at the Fed) and currency in circulation grow with the economy and the federal debt, and that the purchases exist to offset that drain. The balance sheet is not going back down. Whatever the program is called, the printer is no longer in reverse.
Is It QE?
The honest answer is that both sides of that argument hold a real piece of the truth, and the interesting part is what the argument itself signals.
The case for "no" is technical and, on its own terms, sound. QE, as practiced in 2008-2014 and 2020, bought long-dated bonds to push down long-term rates and ease financial conditions: a stimulus tool. Reserve management purchases buy bills, the shortest and most neutral paper there is, to keep the banking system's reserve level "ample": a plumbing tool. The purpose is to offset the mechanical drain from a Treasury cash account that has grown, per Perli, from an assumed $320 billion at quarter-end in 2016 to an average of $850 billion in 2025. No duration is taken out of the market: the Fed is not absorbing the long-term bonds whose prices, and therefore mortgage and corporate borrowing rates, move most. No easing of financial conditions is intended.
The case for "the label is doing work" is historical, and it is uncomfortably specific. In September 2019 the repo market, where banks and funds pawn Treasuries overnight for same-day cash, seized (the overnight rate spiked and the pawnshop briefly ran dry); in October 2019 the Fed announced Treasury bill purchases at roughly $60 billion a month, and Jerome Powell told an economics conference, "This is not QE." The balance sheet grew $316 billion in fourteen weeks. The instrument was the same, the trigger (repo stress) was the same, the insistence was the same, and the episode is remembered today as the moment the post-2018 tightening cycle quietly died. The current program matches it beat for beat, at $40 billion a month instead of $60 billion, with the Standing Repo Facility playing the same canary: its two stress prints, $29.4 billion on October 31 and $31.5 billion on New Year's Eve, bracket exactly the period in which the Fed reversed course.
The June thesis does not need to win the label debate, and that is the point worth restating. The claim was never "the Fed will do stealth stimulus." The claim was about framing: that accommodation, when it comes, arrives under technical names (reserve management, liquidity backstop, market functioning) rather than as an announced change of regime, because the announced version would concede what the fiscal arithmetic implies. Two rate cuts framed as normalization, a QT funeral framed as housekeeping, and $40 billion a month framed as plumbing fit that claim precisely. The framing half of the prediction is behaving. The size half is not, which is the next section.
The Test That Has Not Fired
The scorecard entry is deliberately strict. It resolves only when an episode of 2020 or March-2023 magnitude arrives, and it scores the framing of that episode. For scale, computed from the same weekly series: in 2020 the balance sheet grew $3,010 billion in fifteen weeks. In March 2023, the SVB rescue added $392 billion in two weeks. The current expansion is $189 billion in roughly thirty weeks. On a per-week basis, the 2023 episode ran about thirty times faster. This is not that. Anyone calling the current program "the big print" is reading the label and skipping the arithmetic, and the scorecard entry therefore stays open, unresolved, with a dated interim note.
Honesty also requires the paragraph that hurts the thesis. The June article closed on the claim that bitcoin sits at the bottom of the liquidity gravity well, the default beneficiary when the water rises. Since the balance-sheet trough on December 3, the water has risen by $189 billion, and bitcoin has fallen from $91,345 to $63,044, a 31 percent decline, having already dropped from its October 7 peak of $124,774. Whatever is driving bitcoin's bear market (this website's market-structure work points at forced sellers rather than liquidity), gradual accommodation has visibly not been enough to overcome it. If the thesis pays, it pays on the big print, not on this. Holding the thesis while bitcoin drops through a rising-liquidity regime is exactly the kind of tension a scorecard exists to keep visible.
What the reader takes away is three numbers, all free, all weekly. The balance sheet itself: FRED series WALCL, updated every Thursday from the H.4.1 release, last print $6,724.6 billion; the prediction needs hundreds of billions in weeks, not months, to come alive. Bank reserves: series WRESBAL, last print $2,967 billion against the $2,848 billion October low; a fast break below that floor is how the next stress episode would announce itself. And the Standing Repo Facility: series RPONTSYD, normally near zero; its two spikes above $29 billion marked October's stress and year-end. When those three move together (reserves down, SRF up, balance sheet accelerating), the gradual phase is ending. Until then the honest description of the world is the title. Gradual, so far.