Open the complete story and reading map
The DeFi lessons introduced liquidity, collateral and incentives. The failures of 2022 make a further distinction necessary: a stabilization mechanism can become self-reinforcing in the wrong direction, while an intermediary can misuse assets it controls. Calling both situations simply “crypto failed” makes it harder to identify what evidence a user actually needs.
This article compares mechanisms and selected established facts. It is not a complete account of either collapse, a current litigation tracker or a forecast of creditor recoveries.
Terra: a target price and a feedback problem
Terra's February 2019 protocol explanation described supply adjustments using Luna to absorb changes in demand for its stable currency family. It used TerraSDR as its example. This is evidence of the project's stated design, not independent proof that stability would hold under stress, and not a complete specification of every parameter operating in 2022.
The economic question is straightforward: if a system offers a volatile token in exchange for a supposedly stable claim, who will buy the volatile token when many holders want to exit at once?
Consider a fictional dollar-targeted token S and a volatile token V. In an idealized mechanism, retiring 100 S creates $100 worth of V at the protocol's reference price. At $10 per V, that means 10 V. At $1 per V, the same nominal amount requires 100 V. This arithmetic ignores limits, fees, delays, price-feed problems and halts. It illustrates exposure to the price of the absorber; it is not a historical transaction from Terra.
Possible loss propagation: declining demand for S leads to exits; exits increase the supply of V offered to the market; weaker demand or selling pressure can lower V's price; more V is then needed for the same nominal exit. The link from supply to price depends on market demand. It is a feedback risk, not a mechanical rule that every supply increase causes a specific price decline.
Mechanism analysis does not erase fraud
The SEC's June 13, 2024 release describes UST's May 2022 depeg and reports the April 5, 2024 civil jury finding that Terraform and Do Kwon were liable for securities fraud. It identifies false claims concerning stability and payment use. Explaining reflexivity therefore must not turn the episode into a story of an innocent algorithm alone.
FTX: the account balance and the controlled assets diverged
The U.S. Department of Justice's March 28, 2024 sentencing announcement records Bankman-Fried's conviction and describes misuse of FTX customer funds, including transfers to Alameda. The relevant lesson is about control and truthful accounting, not a failure of Bitcoin's signature-checking algorithm.
Imagine a fictional custodian showing customers a combined balance of 1,000 units. If only 400 units are immediately available and 600 were diverted into risky or inaccessible uses, the dashboard can still display 1,000. A withdrawal request reveals a cash or asset shortfall; it did not create the earlier misuse. This example distinguishes a displayed claim from available controlled assets without asserting that these were FTX's actual balances.
Ask different questions about the two routes
| Question | Stabilization route | Custody route |
|---|---|---|
| What does the holder expect? | A usable exit near a reference value. | Return of assets under the account's terms. |
| What dependency matters? | Collateral, redemption rules, demand and exit-market capacity. | Asset control, segregation, liabilities and authorization. |
| What is insufficient by itself? | A long period trading near the target. | A screenshot of assets without matching liabilities and controls. |
| What stress question helps? | Who absorbs simultaneous exits? | Can this customer withdraw the actual owed asset when many others also ask? |
Changing one component does not finish the risk map
Self-custody can remove a particular intermediary's permission to move your assets, but it does not repair a token's stabilization design. Moving to an exchange with a different matching engine does not establish the quality of its custody. An application can combine a stablecoin, a bridge and a custodian, so both failure paths may matter in the same journey.
Write four lines for any example: the asset you hold, the party or code that controls it, the promised exit and the evidence that supports that exit. Mark missing information as unknown. Do not replace it with a marketing label.
Check your reasoning: would putting a fragile stablecoin in your own wallet guarantee a one-dollar exit?
No. You may change who can authorize transfers while retaining the same asset and market dependencies. Asset-control risk and stabilization risk are separate dimensions, even when they interact.
Continue the risk investigation: Algorithmic stabilization · Custody, settlement and trading after FTX.