SURFACE 2 · SYSTEMIC RISK · NOT THE STAGE DEMO
What bonded credit changes — and what it doesn't.
Bonding governs who can take how much risk and how it unwinds — not what a price is, and not whether code is correct. This page makes both halves of that sentence concrete: the failure classes a bonded credit market structurally excludes, and the ones it does not touch.
Curve / EgorovJul–Aug 2023
~$100M in stablecoin loans (~$110M counting all five venues) backed by 427–460M CRV — about 47% of CRV's circulating supply — as the July 30 Vyper exploit crashed the collateral. Lenders scrambled; the position was too large to liquidate.Delphi Digital · Lookonchain · Aug 1, 2023
EXCLUDED · concentrationIn a bonded credit market this position cannot exist: an enforceable max-size refuses the draw long before one actor's collateral is half the float. A position that cannot be unwound cannot be financed.
Venus / XVSMay 18–19, 2021
XVS spiked ~$76→$144; accounts drew 4,100 BTC + 9,600 ETH against the pumped collateral. Over $200M liquidated; ~$95M bad debt (2,000 BTC + 5,700 ETH) — the protocol's own post-mortem pegged the covered loss at $77M.The Block · The Defiant · Venus post-mortem, May 31 2021
REDUCED · aggregate capAn aggregate budget bounds what any one account can draw against thin collateral — the outsized draw is refused at origination. The price manipulation itself is fence territory: bounds govern authority, not truth.
MakerDAO · Black ThursdayMar 12–13, 2020
Under network congestion, keepers won $8.32M of ETH collateral at zero-DAI bids (1,462 of 3,994 liquidations); 5.67M DAI of bad debt, covered by the emergency MKR debt auction two weeks later.whiterabbit / B.Protocol analysis · Jul 2020
REDUCED · pre-committed unwindA pre-authorized, lender-senior unwind is a contract path armed at origination — it does not auction collateral into a dead mempool and it does not depend on opportunistic keepers showing up. Recovery stays with the principal.
THE HONESTY FENCE — what bonding does NOT help
Bonded primitives do not protect against oracle manipulation (Mango, Oct 2022 — ~$110–116M, MNGO pumped ~13× in 30 minutes; Compound, Nov 26 2020 — $89M liquidated when DAI printed $1.30 on its oracle anchor), venue design failure (Iron Finance, Jun 2021 — the TITAN partial-collateral death spiral, ~$2B wiped), contract exploits (Euler, Mar 2023 — $197M flash-loan exploit, later returned), or market-wide liquidity crunches. Bounds govern authority and unwind order; they do not make prices true or code correct.
The scoped thesis: "Bonded primitives make an agentic market more resilient to authority- and concentration-driven failures — no too-big-to-liquidate actor, no liquidation-cliff discontinuity, no dependence on opportunistic liquidators — converting discrete blowups into pre-committed, solvent-state unwinds. They do NOT improve, and absent staggered triggers and oracle circuit-breakers can worsen, resilience to correlated systemic stress; and they offer no protection against oracle manipulation, venue insolvency, contract exploits, or market-wide liquidity crunches."
Counterfactual-conditional phrasing throughout: "in a bonded credit market, X cannot exist" — a structural claim about enforceable bounds, never a retroactive rescue claim. The stress runs are MODELED and labeled.