Three Reflexive Machines

In roughly 72 hours in May 2022, about $40 billion in market value vanished. TerraUSD, a stablecoin that had held a dollar for months, broke its peg, and the code written to defend that peg became the mechanism that destroyed it. Four years on, the same accusation is thrown at a new instrument: STRC, the $100 "stretch" preferred that Strategy began selling in July 2025, and that critics now call the next Luna.

The resemblance is real, and it deserves to be taken seriously. Strategy's common stock, its STRC preferred, and Terra's UST are all reflexive machines: structures where the price of the thing feeds back into the fundamentals that are supposed to support the price, which feeds back into the price again. George Soros gave the idea its name. Reflexivity is when perception alters the fundamentals it claims merely to observe. All three run on continuous capital inflows tied to a volatile asset, and all three dangle a stability claim to pull those inflows in.

But the interesting work is not noticing the resemblance. It is locating the precise point where the three diverge. The argument here: all three are reflexive and inflow-dependent, yet only one of them carries an automatic, unbounded, code-enforced self-destruct. That single architectural difference separates a death spiral from a corporate-finance squeeze, and conflating the two produces bad analysis in both directions.

To compare them cleanly, hold each up against the same four questions:

1. What is the stability claim? The price the thing is trying to hold, and how hard the promise is.
2. What enforces it? Code, a board, or nothing.
3. What backs it? Exogenous collateral, or a sister asset whose value depends on the same confidence.
4. What happens when confidence reverses? Whether the failure is automatic or discretionary, bounded or unbounded, instant or slow.

Machine One: Terra and Luna, the Pure Algorithmic Case

Terra is the reference design, because it is the cleanest version of the reflexive failure: no discretion, no real collateral, no circuit breaker.

It ran a dual-token system. UST, the stablecoin, was meant to track $1. LUNA was the volatile absorber whose job was to soak up UST's volatility. UST was not backed by dollars, T-bills, or overcollateralized crypto. It was backed by an arbitrage promise against its own sister token.

The protocol guaranteed, by code, that anyone could always swap 1 UST for $1 worth of LUNA, and the reverse, regardless of market prices. That standing arbitrage was the entire peg defense. If UST traded above $1, burning $1 of LUNA to mint and sell 1 UST expanded supply and pushed the price down toward a dollar. If UST traded below $1, buying cheap UST and burning it to mint $1 of LUNA contracted supply and pushed the price back up. In calm conditions it held UST within roughly 1% of a dollar.

Demand for UST was manufactured by Anchor Protocol, which paid about 20% APY on UST deposits. At its peak around April 2022, roughly 75% of all UST supply sat in Anchor, and Anchor's deposits approached $18 billion. That yield was a subsidy paid out of reserves, not income earned from sustainable activity, and it was widely understood at the time to be un-survivable in the medium term. UST became the third-largest stablecoin at about $17.5 billion, and LUNA's market cap exceeded $40 billion at roughly $117 a coin.

In early May 2022, large UST withdrawals from Anchor and a liquidity-pool attack on Curve knocked UST below peg. Then the design did exactly what it was built to do, and that was the catastrophe:

1. UST below peg, so holders rush to redeem 1 UST for "$1 of LUNA."
2. To honor that, the protocol mints enormous amounts of new LUNA.
3. The new LUNA is dumped on the market, and LUNA's price collapses.
4. Falling LUNA destroys confidence in UST's backing, which triggers more redemptions.
5. Return to step two, now worse.

LUNA's supply exploded from about 343 million tokens to roughly 6.53 trillion in about a week. Both tokens went to near-zero. Around $40 to $45 billion in market value evaporated in about 72 hours. Terraform Labs had even bought billions in bitcoin, through the Luna Foundation Guard, as a last-ditch peg-defense reserve. It was overwhelmed almost instantly.

The defining property is that the minting was automatic and unbounded. No board, no chief executive, no human vote could halt it once it started. The peg's defense mechanism was the destruction mechanism. That is the feature to keep in mind for everything that follows.

Machine Two: Strategy, the Corporate Flywheel

Strategy is the common-equity engine, and there is no peg here at all. The reflexive variable is a premium, not a fixed price.

Strategy, formerly MicroStrategy, is the world's largest corporate bitcoin holder, with roughly 845,000 BTC, around 4% of all bitcoin that will ever exist. Its legacy enterprise-software business is now a footnote. The equity is, functionally, a leveraged, actively managed bitcoin-accumulation vehicle wrapped in a public stock.

The key variable is mNAV: the company's market value divided by the dollar value of its bitcoin. When mNAV is above 1, the market pays more than a dollar for each dollar of bitcoin on the balance sheet, and that premium is the fuel. The flywheel turns like this:

1. The stock trades at a premium to bitcoin net asset value.
2. Strategy issues new shares, through at-the-market equity, convertible notes, or preferreds, at that premium.
3. It spends the proceeds on bitcoin.
4. Because the shares were sold above NAV, bitcoin-per-share rises for existing holders. The issuance is accretive, not dilutive.
5. Higher bitcoin-per-share supports the stock, the premium persists, and the loop repeats.

"Bitcoin per share" became the company's explicit north-star metric. At the November 2024 peak, mNAV reached about 3.89x: the market valued the wrapper at nearly four times its bitcoin.

The loop is only accretive while each new share sells for more than the bitcoin it buys. Below 1.0x mNAV, the same action runs backwards: issuing equity now dilutes bitcoin-per-share instead of growing it, and the growth engine becomes a tax on the shareholders it is meant to serve. As bitcoin fell through 2026, that premium compressed: from roughly 2.5x in early 2025 to about 1.16x by spring 2026, and then below 1.0x, a discount. As of mid-June 2026, Strategy's bitcoin was worth about $54 billion against a market cap of about $42 billion, putting basic mNAV near 0.78x. The stock was down roughly 70% over the trailing year, with a reported $12.4 billion loss in the fourth quarter of 2025 and large unrealized losses against a roughly $75,700 average cost basis while bitcoin traded in the low $60,000s.

Why this is not Terra. There is no redemption mechanism. No Strategy share converts into bitcoin on demand, and no holder can force the company to mint and dump anything. The treasury is real and largely unencumbered. The liabilities are long-dated converts and perpetual preferreds, not claims callable on demand. The failure mode is premium compression and a stalled flywheel: painful, slow, and bounded by real assets, not algorithmic hyperinflation.

Reflexivity still bites. Convertible-arbitrage desks short the stock to hedge their bond positions, which can amplify selling on the way down, and the bear case is genuine: the flywheel only creates value while bitcoin trends up and the premium holds. But "stops creating value" is not "self-destructs."

Machine Three: STRC, the Synthetic-Peg Preferred

STRC is the most interesting of the three, because it deliberately manufactures a peg, a soft $100 anchor, using corporate discretion rather than code. It sits structurally between Strategy's common stock and Terra's UST.

STRC, the Variable Rate Series A Perpetual Stretch Preferred Stock, launched in July 2025 at $90 against a $100 par, marketed as a short-duration, high-yield, savings-like instrument. It is one of several Strategy preferreds, alongside STRK, STRF, STRD, and STRE, all of which rank ahead of common equity, and none of which are collateralized by the bitcoin treasury.

The peg mechanism is a discretionary dividend reset, and it is the heart of the comparison. STRC pays a variable dividend, in cash, semi-monthly, that the company resets each month in increments of up to 0.25%, explicitly to keep the share price anchored near $100. If the price drifts below $100, the company raises the dividend to attract buyers and push the price back up. If it drifts above, the company trims the dividend to cool it back down. The rate started at 9% in July 2025 and has been lifted several times; by mid-2026 it sat around 11.5% stated, with the effective yield climbing toward roughly 12.9% as the price fell. Structurally it resembles a money-market peg or a floating-rate note, an instrument that mathematically tends back toward par because the coupon adjusts to whatever the market demands.

STRC plugs directly into Strategy's bitcoin engine. At or above $100, Strategy sells new STRC through an at-the-market program and buys bitcoin with the cash; the peg holding is a precondition for raising money. Issuing at a discount (a $100 face, but only about $90 received) means paying a $100-par dividend on $90 of actual cash, which silently inflates the true cost of capital. Below $100, that program is paused, and the funding channel freezes.

Where the dividend comes from. This is the steelman for the bears. The cash to pay STRC holders comes substantially from new STRC issuance, so new investors' capital helps pay existing investors' dividends. That is the honest description of the cash mechanics, and it is the part that most rhymes with Anchor. Beyond new issuance, Strategy holds a reserve buffer, reported around $1.8 billion, roughly 2.5 years of runway at recent rates, and for the first time since 2022 it has begun selling bitcoin to fund distributions (about 32 BTC sold in late May 2026 for roughly $2.5 million). Selling the underlying asset to service the instrument built on top of it is precisely the buffer-erosion dynamic critics worry about.

In mid-June 2026, STRC closed around $89, an 11% discount to par, its lowest since launch, on heavy volume, with the at-the-market program frozen as a result. The trap analysts describe is this: the longer STRC sits below par, the more Strategy is forced to choose between raising the dividend (a higher cost of capital), issuing common at a discount (dilutive), or selling bitcoin (eroding the collateral and the mNAV premium).

The Same Four Questions, Side by Side

Reflexivity Is Not a Death Spiral

Everyone in this debate is right about the resemblance. The question is whether the resemblance is load-bearing. Here is the precise decomposition.

What the three genuinely share:

• A capital-inflow cycle tied to a volatile asset, bitcoin or LUNA.
• A high yield (STRC, Anchor) or a high premium (Strategy) used to attract that capital.
• Real stress when the cycle reverses, with each instrument's "stability" becoming the pressure point.

What separates Terra from the other two is the circuit breaker. Terra's peg defense and its self-destruct were the same code path. Once UST broke peg, defending it required minting LUNA without limit, and minting LUNA without limit destroyed UST's backing. There was no human in the loop, no discretion, no off-switch, and no floor. LUNA could be minted to infinity and its price driven to zero.

STRC and Strategy's common stock have none of those four properties:

• Discretion, not automation. A board sets the STRC dividend and can raise it, hold it, cut it, or suspend it, and it can pause the ATM. Nothing mints automatically.
• Real, unencumbered collateral. Both are backed by an actual bitcoin treasury that exists independent of the instrument's price. UST was backed by faith in a token that shared its fate.
• No redemption into the volatile asset. No STRC or Strategy holder can force conversion into bitcoin. The mechanical engine of Terra's spiral, forced on-demand minting of the absorber, does not exist here.
• A seniority waterfall, so losses are bounded. Preferreds rank ahead of common, and the treasury is large and largely unpledged. The worst plausible STRC outcome is painful but bounded (suspended or accruing dividends, a sustained discount, dilution of common), not zero.

Terra had a forced, unbounded, instantaneous failure path. Strategy and STRC have a discretionary, bounded, slow one. That is a difference in kind, not degree.

Where the Bears Are Right

Dismissing the analogy entirely is also a mistake. The legitimate concerns:

• New money pays old money. STRC dividends are substantially funded by fresh STRC sales. While the ATM runs at par, that is sustainable; if it freezes for long, the reserve drains and the structure leans on bitcoin sales.
• Selling the collateral to service the claim. The May 2026 bitcoin sales, the first since 2022, and the use of a buffer originally built to avoid selling BTC, show the risk shifting rather than disappearing.
• A reflexive attack surface. A bear can short bitcoin to seed fear of a cash crunch, STRC discounts, the ATM freezes, the market anticipates forced bitcoin sales, and more fear follows. That is genuinely Soros-style reflexivity, even without an automatic minting trigger.
• Growing fixed obligations against falling collateral. Each STRC issuance adds a perpetual dividend obligation that must be serviced indefinitely, even as bitcoin and mNAV fall.
• Convertible-arb amplification on the Strategy side accelerates drawdowns mechanically.

The honest framing several analysts have landed on is that the fear is mostly a bitcoin story, not an STRC story. If bitcoin finds a floor, the discount and the dividend fears resolve and the floating coupon pulls STRC back toward par. If bitcoin keeps falling for years, the squeeze grinds on regardless of how clever the wrapper is.

Synthesis

The three sit as points on a single axis, defined by how hard the stability promise is and what enforces it, running from pure market sentiment to pure code:

• Strategy common makes no promise. It carries a premium the market grants or withdraws. Reflexive, fully discretionary, bounded by a real asset floor.
• STRC makes a soft promise, near $100, enforced by a board's monthly dividend dial. Reflexive, discretionary, bounded by seniority and real collateral, but with Anchor-like cash mechanics and a real freeze-and-squeeze failure path.
• Terra's UST made a hard promise, $1, enforced by automatic code with no collateral and an on-demand mint into a reflexive sister token. Reflexive, non-discretionary, unbounded, which is why it went to zero in three days.

The resemblance everyone points to is real. Three machines that run on confidence-driven inflows around a volatile asset, each paying up in yield or premium to keep the inflows coming, each stressed when the cycle reverses. But the thing that killed Terra, a forced, brakeless, unbounded minting loop with no exogenous backing, is the one ingredient Strategy and STRC structurally lack. They can be badly hurt by a long bitcoin bear market. They cannot, by their own architecture, hyperinflate themselves to zero in 72 hours.

"The next Luna" is the wrong frame. "A reflexive, leveraged bitcoin structure with a discretionary brake and a real asset floor, under stress in a bitcoin downturn" is the right one. Less dramatic, and considerably more useful for sizing the risk.

This is mechanism analysis, not investment advice. The figures (bitcoin price, mNAV, STRC price, dividend rate) move daily and are external to this site's pipeline; re-check the live numbers before relying on any of them.