Mainnet Position Proof

Ledger reconstruction of real Aave V3 positions, from real mainnet data

Credit decisions need history that lives on another chain. The obvious way to get it is to sum a protocol's events. We measured that approach against reality and it is wrong by up to 66.8084% on a single position. This page is the measurement, the replacement primitive, and the proof that the replacement reconciles.

01The finding: event summing understates a position

A real mainnet borrower with 311 pool events. The naive method sums Supply minus Withdraw per reserve. The right column is the same position read from protocol state at the same block.

02The replacement: the token's own ledger, anchored at a proven zero

The primitive is the aToken's Transfer ledger, not the pool's events. A ledger sum is only meaningful if you know where it started, so the engine finds the most recent block where balanceOf returns exactly zero and anchors there, which makes coverage complete by construction rather than by assumption.

40 wallets, positions from 50.0204 to 0.0000 in aEthWETH. 1 matched exactly and 38 of 40 landed within 10 bps, with a largest residual of 4.36 bps.

03Parity: every topic pinned against the chain, with negative controls

Constants that are wrong fail silently, so each declared signature is checked against real logs in a 10,000 block window and paired with a negative control that must return zero for the method to be credible. 9 of 10 signatures confirmed live.

04End to end: the same flow the contract runs

The engine then values the position through attested prices. Executed across 10 transactions on local anvil 31337 using the real mainnet inputs above, then read back from the deployed contracts rather than from the script's own console output.

05And the number it produces

Proving a position is only worth anything if something consumes it. The same broadcast deploys a policy layer that sizes a credit limit from the proven net worth, read back off chain with the rest of the stack.

06Read live from Creditcoin testnet, no wallet required

The engine above is now deployed. Every number below is read from those contracts in your browser as the page loads, including the bytecode check, so each one can be pasted into Blockscout and confirmed independently. Nothing on this panel is cached or filled in.

07A verified record, consumed by someone else

The standing registry publishes a proof-anchored record any protocol can read. Until this deployment, nothing read it, which made portability an interface rather than a fact. So a second, independent product was deployed against it: a merchant that defers payment for an item, using its own policy, on a record it did not create. Spark cannot change that policy, and the merchant cannot change what the record says.

08What is not true yet

• The generation-2 stack has now been used end to end once: a 0.0199 ETH Sepolia balance attestation opened a 0.00398 ETH limit with no deposit, and the full limit was then drawn from the product itself. That is the first time this path has carried a borrower from proof to credit, and it is still one account. A mechanism that works once is not demand, and one draw is not a credit book.
• Interest is never fabricated from a timestamp or a rate. The only path that moves a position ahead of the ledger is an attested state balance, and the residual is capped and reverts past the cap.
• The sample is aEthWETH only and biased toward recent depositors. Morpho WithdrawCollateral has no observed logs in the window, so that signature is unconfirmed.
• The attestor is trusted to submit already verified values rather than the contract calling the precompile directly. That trust boundary is documented in the threat model.
• The credit limit has no on-chain enforcement path yet. Nothing draws against it, so it is a policy output rather than a funded line.
• The portability pair is exercised by one account acting as both borrower and merchant. It shows the mechanism is consumable by a different contract under a different policy; it does not show two parties using it.

Spark's live product is separate and unaffected. Paying on Sepolia and opening credit on Creditcoin with dual Attestcoin proofs is deployed and working today, with on-chain history you can read on Blockscout.

cd app && TARGET=40 DISCOVERY_CHUNKS=8 node scripts/position-scale.mjs   # the 40-wallet corpus
#   the bare command defaults to TARGET=8 DISCOVERY_CHUNKS=1 and reconstructs 8 wallets, not 40
cd app && node scripts/protocol-topics.mjs     # topic parity + controls
cd app && node scripts/gen-evidence-module.mjs # regenerate this page's data
cd contracts && forge test                     # 518 tests
cd contracts && bash script/deploy-all-cc3.sh  # the CC3 broadcast (dry run by default)
cd contracts && bash script/standing-consumer-cc3.sh  # the portability pair (dry run by default)