Notes on markets and mechanisms

By share price LLP returned 17.2% annualised over 90 days and 7.8% over a year that included the 10 Oct 2025 crash, but entry requires holding a third of the deposit in staked LIT and the rollup's contracts can be upgraded with no delay.

Zero fees and isolated collateral, but the OLP market-making vault is closed to outside capital, unrealised profit is an unsecured claim on OLP, and ETH perp funding is a third of Hyperliquid's.

Two-thirds of NNE's market cap is cash, so the market pays ≈ $270m for all the reactors combined; the key catalyst — an NRC construction permit — is expected only in 2027, and dilution and fuel supply are the main risks until then.

USDC reserves cover 100.3% of supply with only 9.5% in banks, but the likeliest way to lose access in DeFi is Circle blacklisting an entire pool contract over someone else's court case.

There are few extra coins outside Deribit and all of them are short-dated; what Deribit really lacks is options on spot ETFs (IBIT, ETHA, BSOL) running to 2029 and cleared by the OCC — longer tenor and the safest payout.

The liquid crypto options universe is BTC and ETH; implied vol on Deribit, OKX and Bybit agrees within a point, and alts with more than a month to expiry exist in practice only on Deribit, capped at ~4.7 months by the exchange's own listing rule.

Derive marks options almost exactly like Deribit, but it is a weaker counterparty, its alt order books are empty, its yield-bearing collateral pays less than Deribit does on the same assets, and the V3 migration splits the insurance fund four ways.

The AUTO token really accrues ≈8% a year, not the 24.77% on the banner; looping to 4.5× gets up to 17.6% but liquidates on a 2.5% price drop, and the oracle price is a model accrual that will be marked down in a single step.

BTC's MVRV peaks fall cycle after cycle and pre-top funding fires only every other time — neither is a reliable exit signal; both work only as a phase marker or a confirmation.

Lighter's pre-IPO OURA perp traded 14.5% above the top of the IPO range, its price is computed by the venue from a thin book, and post-listing settlement is at the venue's discretion.

Freedom24's own terms let it hold client securities in the company's name and outside the EU, and the SEC has sent the group a Wells Notice — fine for unleveraged listed securities, not for size.

Client securities are protected by SEC Rule 15c3-3 segregation and a $22.3bn equity base, not by insurance — the Lloyd's policy's aggregate cap is 0.016% of client assets.

Lido's core hasn't been exploited in six years and the withdrawal queue clears in about two days; the real risk has migrated to L2 bridges and products built on top of Lido.

USDC on Monad is native Circle issuance via CCTP, while ETH is an IOU from a ten-month-old Wormhole + Axelar bridge — so chain risk has to be assessed asset by asset.

PT locks the rate in sUSDai terms, but offers no protection against losses in a young GPU loan book that the lender marks itself.

Uniswap V3's core has gone unexploited for over five years and nobody can change it, the protocol fee is already at its cap — LP money has been lost to phishing and third-party position managers, not to the core.

The largest crypto options exchange has never socialized a loss in eleven years, but on 1 September 2026 it dropped its daily proof of reserves, and its rulebook explicitly allows non-disputable socialized losses.

Hyperliquid's engine came through the largest liquidation cascade in crypto with no bad debt — by force-closing winning positions — and its emergency rules are effectively controlled by the exchange's own foundation.

Three years without an exploit and zero bad debt on 10 Oct 2025, but depositor exits are capped by idle reserve liquidity and two multisigs with 12–24h timelocks control the rules.

Royco's senior tranche on savUSD lifts the Avant pool-loss threshold from 14.28% to 34.85% for ≈ 21bp a year, but the tranche logic has no published source code and the docs describe a different deployment.

A Solana memecoin launchpad quoted against tokenised stocks: the "dividends" are a transfer tax between holders, 88% of all fees landed in 13 days, and the 16.4% supply burn would cost four times as much at today's price.

Liquidations work and the supply index never dipped, yet Aave's pools carry $213m of recorded, unallocated deficit — and withdrawable USDC on Ethereum drops almost to zero every night.

Since 7 Aug 2026 weETH is plain staked ETH: over a year it earned exactly what wstETH did, its redemption queue survived a real stress test, but market exit is capped at 0.63% of supply and two thirds of the token sits in Aave.

A token on a stake in a fund that hasn't yet bought any Neuralink stock is sold at a $44bn valuation — 4.6x the only real round — and the markup eats a third of the upside in the platform's own scenario.

Lighter's ANTHROPIC perp prices off its own thin order book, already implies ~$1.93tn — double the last real round — and the venue can rewrite IPO settlement terms on a day's notice.

Robinhood's SPY token is not an ETF share but a Jersey debt note whose issuer can freeze and confiscate balances; as a substitute for SPY at a broker it loses on every count except 24/7 trading, and its LP yield depends heavily on the measurement window.

The three largest tokenized stock programmes sell offshore SPV debt, lose to a broker on ownership and win only on access — while the asset class converges on the regulated US rail being built by DTCC, Nasdaq and the SEC.

Avant's ETH junior tranche carries the highest leverage of its peers (8.5x) on a 0.313% reserve, and 47% of the product is the same protocol's senior dollar tranche — itself three quarters USDe.

Avant's junior tranche on Avalanche runs 7.4x leverage over a reserve of 0.82% of the pool, three quarters of assets sit in USDe, an Aave–Morpho loop underpins both tranches, and the seven-day cooldown carries full risk with zero yield.

Three of the four gold wrappers are claims on real metal and the fourth has no metal at all; in a crisis each protects against something different, and organic DeFi yield on tokenized gold is essentially zero.

The basis trade is now just 13% of USDe backing, and 37 bp over T-bills doesn't pay for the credit and counterparty risk you're taking.

Strata's junior tranche on sUSDe runs 7.9x leverage for 94bp of headroom to zero, redemptions only lock after the tranche has lost 65% — and four addresses alone cover that threshold.