Treasury published its borrowing plan on 5 August and told the market it expected to end September holding $950 billion in cash. It closed above that number on 6 August. The next day.

It has not been back below since. The account closed Monday at $1.0236 trillion, the first time it has ever finished a day above a trillion dollars.

Money moving into the Treasury's account at the Fed leaves the banking system. That is not a reading of the data, it is an accounting identity. The Treasury General Account and bank reserves are both liabilities of the same balance sheet, and one grows at the other's expense.

For two years that did not hurt, because the cash came out of somewhere nobody needed it. The Fed's reverse repo facility held $2.55 trillion at the end of 2022, a parking lot for money market funds with nothing better to do. Treasury issued bills, the funds bought them, the parking lot drained, and bank reserves were barely touched.

On Tuesday the parking lot held $725 million. Not billion.

One caveat we would rather hand you than have you find. The facility took $6.7 billion on 31 August, a month-end spike of the sort that appears every quarter when funds tidy their books, and was back under a billion the next day. It still works. There is simply almost nothing parked in it on an ordinary day, and that is the part that matters for what comes next.

Because Treasury is not finished. Its own projection puts the account peaking near $1.05 trillion in late October, give or take $50 billion.

Here is what does not fit, and we would rather show it than bury it. Credit is pricing none of this. High yield spreads closed August at 2.63%, twenty two basis points tighter over the month and inside the tightest 2% of readings in the available series. Reserves draining into a possible rate hike should show up in credit before it shows up anywhere else. It has not shown up at all. Either the plumbing story is wrong or the credit market is, and we do not know which yet.

Three metrics.

Metric 1, Fear and Greed: 63. Still Greed. The last five prints run 68, 69, 62, 69, 63, which is not a trend in either direction. Sentiment has gone sideways while the tape has not, and we would rather say that than draw a line through noise.

Metric 2, ETF flows: August closed above $3 billion. The strongest month of 2026, against $1.97 billion in April. One number complicates the headline. In the week to 28 August the funds took in $924.5 million net while BlackRock's IBIT alone took $938.3 million, so the rest of the complex netted to roughly minus $14 million between them. Not uniformly: Fidelity's FBTC and Morgan Stanley took money in, while ARKB and GBTC gave more back. A record month and a one issuer bid are not the same thing.

Metric 3, BTC dominance: 59.1%. Bitcoin was $77,541 on Tuesday evening against a $2.62 trillion total crypto market. This is also where Strategy lives, and Strategy bought bitcoin on Monday for the first time since June. 4,603 coins at an average $80,318, which is above where the coin trades now. What almost every write-up called a ten week pause was nothing of the kind, and the real sequence is in RADAR.

That purchase should also have settled one of our dated tests. It did not, and the reason is our fault rather than the market's. The STACK has it, in full.

📣 THIS SPOT IS OPEN

Baseline Crypto reaches nearly 7,000 Bitcoin-focused readers every Monday and Wednesday morning, opening just under half of what we send. Direct sponsor placements are open, premium position, top of the issue, one per send.

Reply to this email for rates.

This is the standard answer and it is not a stupid one. The TGA is a checking account, not a vault. Treasury collects taxes and issues debt, then spends it on payments that land back in the banking system within weeks. Over a long enough window the drain reverses itself. Anyone who panicked about the 2023 rebuild and shorted into it lost money, and the neutrality argument has the better track record of the two.

The flaw is that it assumes the balance is set by cash needs. It is set by career risk.

Ask what the person choosing the number is paid to avoid. A Treasury Secretary who runs the account too low and has to prioritise payments during a debt ceiling fight owns a historic failure with his name permanently attached. One who holds an extra hundred billion owns nothing at all. That cost is diffuse, unattributed, and paid by bank reserves rather than by him. The penalty is entirely one sided, so the target only ratchets in one direction. Before 2015 a normal balance was under $100 billion. The published assumption for this September is $950 billion.

Then take the buyback story in RADAR, and notice what it implies even at second hand. Officials inside Treasury are reportedly weighing whether to point TGA cash at buybacks of long dated debt. Nobody weighing that is planning to spend the balance back down into bank reserves. That is a building looking at a trillion dollars of idle cash and asking what it could do to the long end. The sourcing is thin and we are not building a forecast on it. We are pointing at the incentive it reveals, which does not depend on the story being confirmed.

Which is the answer to the strawman. The money does come back. It does not come back the way it left. It returns as bond buybacks, or as a permanently larger buffer that nobody wants to be the official who shrank, and both of those leave reserves lower than the identity implies.

What we are not claiming. We are not saying reserves are about to go scarce and we are not calling a funding event. When the plumbing genuinely broke in September 2019 the tell was repo rates spiking, not a large TGA. Repo is calm. Credit is calm. If reserves were biting you would see it in one of those two places first, and we would rather say that plainly than sell you a crisis built on one line going up.

New dated test, grades Monday 5 October on the Daily Treasury Statement for 30 September 2026, published the following business day. Branch A: $1,050 billion or above, meaning Treasury reached its own projected late October peak a month early and the drain is running ahead of plan. Branch B: $950 billion or above but below $1,050 billion, on plan. Branch C: below $950 billion, meaning Treasury undershot the assumption it published on 5 August. One published number, three ranges, no gap between them and nothing outside them. After the STACK below, we owed you a test built that way.

Three numbers that define the next six weeks.

$1.05 trillion, Treasury's own projected peak. Late October, plus or minus $50 billion, published by Treasury in the 5 August refunding statement rather than estimated by us. The band matters more than the point, so here is the whole band: it runs from $1.00 to $1.10 trillion, and the bottom of that range sits below where the account already closed on Monday. Read literally, Treasury is telling you the peak might be another $76 billion away or might already be behind us. On the central case it is another $26 billion out of the system. We are taking the central case and showing you the low branch, because a projection quoted without its error bars is just a number with better marketing.

2.63%, high yield spreads, and the number that falsifies us. If the reserve story in this issue is right, credit is where it shows up first, because leveraged borrowers feel funding stress before anyone else does. Spreads are instead twenty two basis points tighter on the month. We are not going to explain that away, so here is the commitment, with the middle filled in rather than left as a convenient gap. At or inside 2.70% when the TGA peaks, our framing was wrong and we will say so in those words. Between 2.70% and 3.00%, the signal is ambiguous and we will report it as unresolved rather than claim it. At or through 3.00%, the drain is doing what we think it is doing.

16 September, the FOMC decision. Polymarket has a 25 basis point hike at 57.5% against 40.5% for no change. A hike into a reserve drain is the tightening that actually bites, because it raises the cost of funding at the same moment the quantity of it is falling. The meeting matters less for the decision than for what Warsh says about the balance sheet alongside it, and almost nobody is positioned for that half of the statement.

Forget analyst predictions. Polymarket is a real-money prediction market, where traders put actual dollars on outcomes. Scorecard first, then the board. Not financial advice. Our read. Disclosure: we hold personal positions in Polymarket itself and may earn a commission from Polymarket referrals. These markets are thin and the odds below move, so read direction over ticks.

Scorecard: the test we set on 26 August, due to grade 23 September, is VOID. Second time in three weeks we have written branches that do not cover the outcomes between them.

It asked what Strategy would do with its cash pile. Branch A required a bitcoin purchase disclosed in a weekly 8-K, of any size, funded from the USD Cash account. Branch B required no purchase at all through 21 September, with USD Cash holding at or above $1.0 billion.

Monday's filing did neither. Strategy bought 4,603 coins, which kills Branch B outright. It funded them from a $602.8 million share sale, stated in the 8-K's own footnote. Branch A is therefore unmet. The USD Cash account was never drawn on. It went up, by $30.0 million, to $1.61 billion.

Neither branch resolves. Recorded VOID, not a pass, and not quietly dropped.

On Friday we wrote that the test was tracking toward Branch B, and Monday's issue said it again on the morning the 8-K landed. Both wrong, and that is the smaller error. The larger one is that we wrote this test on 26 August, in the very issue where we voided the previous test for this exact defect, and signed it off with the line note that these two branches now cover the outcomes between them, which is more than we managed last time. They did not. We took a victory lap for fixing a flaw we had just reproduced.

The root cause is narrow enough to name. Both broken tests specified a mechanism, then quietly assumed it had two states. A company can buy with cash on hand, with fresh equity, or with debt. We wrote branches for one of those and called the set exhaustive. The 17 August test made the identical mistake with a price level and a sale.

So the standard tightens again, structurally rather than as a promise to be careful. Dated tests from here resolve against a single published number cut into ranges that leave no gap, which is how the TGA test in the FILTER is built. A number line has nowhere to hide. Void a third and the honest conclusion is that we should not be writing tests about corporate behaviour at all, and we will say that too.

The September hike we declined on Monday is now 57.5%. We passed at 52.5 and said why. It has since added five points, which on paper is money we did not make. The reasoning stands and we are not revising it after the fact: we declined because the repricing had already happened on public information, not because we expected a reversal. Being right about process and poorer for it is the ordinary case, not a defect. A newsletter that quietly rewrites its reasoning when a passed trade goes on to win is not worth reading.

Two contracts come off the board for space, and both keep their published re-entry conditions. CLARITY, exited at 41% and last quoted 12.5%: Congress returned this week, which is the catalyst we named, and nothing has cleared markup, so the condition is unmet. Zero Fed cuts in 2026, exited at 88.75% and last quoted 87.9%: unchanged through the entire September repricing. Neither is being dropped because it stopped mattering, and we will report either the moment it triggers.

The book is flat for the ninth consecutive issue, and Monday's issue miscounted it. We printed "seventh" on Monday having already printed "seventh" on Friday. The run reads fifth on 19 August, sixth on 26 August, seventh on 28 August, eighth on 31 August, ninth today. Corrected here, because a scorecard that cannot count its own idle streak has no standing to grade anything else. The entry condition stands: the next position comes from the reserve drain thesis or from nothing, and if high yield spreads widen through 3.00% into the October peak, that is the trade, published here before it is taken.

Fed decision in September, 25 bps increase | Market: 57.5% ⚪ NO TRADE, passed at 52.5 and standing by it
Up from 52.5% on Monday, with no change at 40.5% and cuts at 0.6% combined, on $77.3 million of volume. What would put us in: a fall back under 40% with Warsh's position unchanged, which would mean the market disagreeing with a Fed Chair who has now made the same argument twice inside a week. That is a mispricing claim, and it is the thing our 26 August threshold was not.

Fed rate hike by the October meeting | Market: 66.5% 🟡 WATCHING, the more honest contract of the two
September sits at 58% here against 66.5% for October, so the market prices roughly nine points on the Fed waiting one meeting and moving anyway. That gap is the real question and it is better priced than the September leg. It is also the one that overlaps our thesis, because late October is when Treasury's own projection puts the cash peak. What would move it: any FOMC language about the balance sheet on 16 September.

What price will Bitcoin hit in September | Market: 77.5% a dip to $75,000, 69.5% a touch of $80,000 🟡 NO TRADE, the market is pricing both legs
Those add to well over 100 because they are not exclusive, and the market is saying September is a wide month in both directions. It has a point already: the month was hours old when bitcoin printed $76,398 and the $77,500 leg resolved yes. A volatility read, and we have no edge on volatility. What would make it interesting: the $75,000 leg holding above 80% while spreads are still tight, which would mean the dip is being priced by positioning rather than by anything in the macro.

When will Bitcoin hit $100k | Market: 58.5% by 30 September 2027, 84% by 31 December 2027 ⚪ NO TRADE, and thin enough to say so
A twenty five point spread between September and December 2027 is a market with no view on timing at all. Read it as long horizon sentiment and nothing more, and discount it further because the whole event has traded $37,000 and the two legs quoted here hold four figures between them. We hold no position and would need a horizon under six months to have one.

Track the whole board live at polymarket.com, free, no account required.

Monday's question got overtaken by a filing.

We asked whether the remaining 47.5% on the September hike was real disagreement or process risk. Before anyone could argue it, Strategy filed an 8-K that landed outside the branches we had written for it, for the second time in three weeks. That is not the filing being clever. That is us writing tests that assume the world has two options.

So this issue's question is the one we genuinely cannot resolve.

Treasury's own projection has the account peaking anywhere between $1.00 and $1.10 trillion by late October, which means the drain either finished last week or has another $76 billion to run. High yield credit is twenty two basis points tighter on the month and pricing none of it. If both are still true on 31 October, which one do you stop believing?

Our answer is in BEYOND THE CHARTS and it is falsifiable on purpose. Spreads inside 2.70% at the peak and we were wrong about the mechanism. Through 3.00% and we were early rather than wrong. There is a third possibility we like least, which is that reserves stopped being the binding constraint years ago and everyone still writing about net liquidity, us included, is reading a gauge that came unplugged.

If you think it is the third one, make that case. It is the argument most likely to change how we cover this, and the one we are least equipped to make ourselves.

Hit reply. We read every response, and the best answers run Monday.

See you Monday, with the first September ETF prints, whether Strategy adds again and out of which pocket, and the FOMC two days out.

The Baseline Crypto Team

Three tools we built because we wanted them and could not find them. Free, no signup, no accounts, and nothing you type ever leaves your browser.

Is your hardware wallet screwed? Check your device and firmware against every logged advisory, including the ones where updating your firmware does not save you, because the seed itself is already weak.

How fragile is your self-custody? Nine questions, scored out of 27. Every question exists because of a documented loss.

Our track record, losses included. The public ledger of every dated call. It is running behind the last few issues and we are fixing that this week, so today the STACK above is the current record and the page is the archive.

Strategy sold at $64,262 in August and bought back at $80,318 on Monday. If you have a friend who did the same thing in miniature, selling into the lows and buying the rally on record-month ETF headlines, they are in that trade with none of the balance sheet behind it. Send them the custody audit, because the thing that turns a bad entry into a total loss is not timing.