Three Billion Dollars of Shorts Died This Week. The Money Supply Did Not Move.

Bitcoin has risen about 25 percent since Wednesday and traded past $78,000 by Saturday morning in Asia (CoinDesk, August 22, 2026). Ether rose 18 percent in the first leg. XRP rose more than 15. Solana came along.

The mechanism was specific, and small. The US Treasury said it would double the size of its buyback operations in the longest dated government bonds, to $4 billion from $2 billion per operation. The 30 year yield fell from 5.34 percent, a 19 year high, to around 5.19 percent. That was the whole intervention (CoinDesk, August 22, 2026).

What it set off was a liquidation. Nearly $3 billion of positions closed out on Thursday alone, across 172,108 traders, with shorts roughly 92 percent of the total, which CoinGlass records as the largest wave of forced short closures going back to 2021 (CoinDesk, August 20, 2026). Across Thursday and Friday together the figure was around $4 billion (CoinDesk, August 22, 2026).

There was a policy story running alongside it. The President pushed Congress to advance the Digital Asset Market Clarity Act at a White House event attended by executives from Coinbase, Gemini, Ripple and Chainlink Labs, with a procedural vote scheduled for September 15 (CoinDesk, August 19, 2026). The CFTC chairman told the inaugural Innovation Advisory Committee that his agency would not sit idle, putting staff on notice to write crypto rules if the Clarity Act fails (CoinDesk, August 20, 2026). US spot bitcoin ETFs took in $517 million on August 19 and ether ETFs took $189 million (CoinDesk citing SoSoValue, August 20, 2026).

So: a liquidity trigger, a policy catalyst, record short pain, and real inflows.

We ran a committee session into the middle of it.

The number nobody was looking at

Our fact sheet carries a line most market commentary does not, because it is boring and it does not move much. Aggregate stablecoin supply.

On August 21, the sheet the four seats read put it at $309,249,862,213, down 0.11 percent over thirty days (DefiLlama, August 21, 2026).

Refreshed for publication a day later, it stood at $310,070,872,021, up 0.27 percent over thirty days (DefiLlama, August 22, 2026).

That is the whole of it. Bitcoin added roughly a quarter of its value in three sessions. The total supply of dollars living on chain grew by about $821 million across the same window, and is up less than half a percent on the month.

Stablecoin supply is the most direct measure available of money arriving in crypto and staying there. It is what gets minted when someone converts dollars into something they can spend on chain. When it expands, new capital has entered. When it is flat, the money moving prices is money that was already inside.

Investor C, the seat that reasons from macro and reflexivity, put it in one line:

This rally is leverage and short-covering, not new money arriving.

That is checkable, it is sourced, and it reframes the week. A record short liquidation is not evidence of new demand. It is evidence that people who were positioned one way were forced to trade the other way. The buying was real and it was compulsory.

None of which means the move is fake or that it stops. ETF flows enter through channels that never touch a stablecoin, and Investor C said so against its own argument, noting that $517 million of single day bitcoin ETF inflows shows capital can arrive through traditional rails without appearing in on chain money supply at all.

By the weekend the publisher that reported the rally had reached the same description of it, calling the week a squeeze led rally in its own review (CoinDesk, August 22, 2026).

But it does change what the week is evidence of.

What we put in front of the committee

Jupiter, the largest trading and routing layer on Solana, and one of the few tokens in this universe attached to a business that collects real fees.

The sheet the four seats received carried, among other things:

All figures from DefiLlama, CoinGecko and Helius, each carrying its own url and date on the sheet the seats read.

On the face of it that is a cheap business. A category leader at 1.9 times market capitalisation to annualised fees is not a valuation that usually needs an argument.

All four seats found the same hole

The four investors run on four different models, receive one identical sheet, have no tools, and never see each other's work. They returned four different words, which is the ordinary outcome here.

Underneath the disagreement, all four independently attacked the same figure, and it was the figure that made the protocol look cheap.

Investor A, quality and long horizon: "The fee window is suspect. Fee run rate is up 21% 7d versus 30d, and the review period contained a violent short squeeze. Fees earned in that tape are not a clean run rate."

Investor B, deep value: "that environment may temporarily inflate trading-related fees."

Investor C, macro: "The fee acceleration could be a function of the squeeze itself, not organic demand. Annualizing fees earned during a record liquidation event will overstate the run rate."

Investor D, growth, and the only buyer in the room, listed it as the thing that breaks its own case: "my thesis fails if current annualized fees of $339,326,658 reflect a temporary cyclical peak driven by short-term market liquidation volume rather than durable organic demand."

Four seats, four philosophies, four different votes, and one shared conclusion that the 21 percent fee acceleration is exactly what a record liquidation event would manufacture. The seat that voted Buy named it as its own thesis breaker.

That is a more useful output than the verdicts. A committee that agrees on the rating tells you very little. A committee that disagrees on the rating and converges on which number is contaminated has told you where to look.

Three of the four also refused to get past the same three blank lines. The unlock schedule, the holder concentration and the protocol treasury are all not available on our sheet, and Investor B would not price the asset at all because of it: "Intrinsic value cannot be estimated." That is the same reason it gave when this protocol was first reviewed on August 9, and it has not moved.

The committee did not move

This protocol went in front of the same four seats on August 9, twelve days earlier, on a different sheet, before the Lend v2 launch, and before any of this week's move.

Four identical words. Three convictions up by one. One unchanged.

The seats were not shown the earlier session. They were told the protocol had been reviewed before and instructed to reason afresh from the sheet in front of them, which is the same instruction every session carries.

We publish this comparison because we recently published its opposite. On August 12 we ran three companies through this committee twice, twenty five minutes apart, on identical sheets, and seven of twelve seat verdicts came back different. We said at the time that run to run variance is a real property of a committee made of reasoners.

Both things are now on the record, and they are not in conflict. Identical inputs twenty five minutes apart produced movement. Genuinely different inputs twelve days apart, across a product launch and the largest short squeeze since 2021, produced none. The noise sits inside a single reading. The signal, so far as we can see it, sits in whether anything actually changed about the business.

For Jupiter, on our sheet, nothing did. The fees went up in a week when fees were always going to go up. The three disclosures that would settle the argument are still missing. Lend deposits and active loans both slipped over the month, which every seat noticed.

What this note does not establish

It does not score anything. No verdict in our record has been scored against an outcome, and this session is now one more unscored row rather than a result.

It does not forecast a price, for bitcoin or for anything else. No seat was asked for one and none gave one. The stablecoin observation is a statement about what has already been measured, not a claim about what happens next.

It does not say the rally is wrong. It says the rally was funded by positioning rather than by new on chain money, on the one measure we can source, and that ETF flows are a real channel that measure does not capture.

It is not a recommendation to buy or sell anything, and BagsCapital holds no position in JUP.

A note on timing. The committee session ran on August 21 and every figure on the sheet the seats read carries that date. The market data above was refreshed on August 22 before publication, and both dates are marked wherever they appear, because a fee figure and a price figure measured a day apart should never be presented as though they were measured together.

One methodology note, published because it cost us time. Our token sheet builder resolves a mint address, and the obvious address for Solana itself returns Wrapped SOL, the wrapper contract, carrying a circulating supply of 11,978,496 and a market capitalisation near $1.07 billion. Those are real figures for the wrapper and they are useless as a description of the network. We caught it before the session rather than after, and the session you have just read is a protocol with a fee engine behind it for exactly that reason.