Solutions / Onchain Credit
Prove every loan met your committed criteria, so no allocator has to re-underwrite you first. The model and the borrower file never leave.
The problem
$20B of onchain credit is outstanding, and every dollar of it was underwritten twice: once by you, then again by whoever funded it.
Capital arrives at the speed of relationships and data-room access, and none of that work carries to the next one.
It is the reason your book performs, and it is the one asset you cannot show to the people asking why it performs.
Borrower data is usually restricted by contract or by law, so transparency is not a lever available to you even when you would pull it.
The boundary
Underwriting is the product, and it is made of exactly the material you cannot publish. A proof carries the conclusion across without the evidence.
Never leaves your infrastructure
Borrower identity and financials
Names, statements, covenants and the diligence file never enter the proof as anything but private witness.
The underwriting model
Weights, thresholds and the feature set that make your credit calls yours stay on your infrastructure.
Pricing and terms per loan
The spread you struck, and why, stays commercial information.
Goes onchain
A proof, verified onchain, that this loan was originated by the committed model against the committed criteria. Nothing else crosses.
256 bytes, verified for roughly 300k gas. The cost is the same whether the model behind it is a threshold rule or a scorecard with a thousand features.
What POY enforces
Origination rules compiled into the program the pool proves against. A loan outside the credit box is not flagged for review. It is not written.
Eligibility, concentration limits, tenor bounds and collateral requirements are enforced at the moment a loan is written.
The proof covers the process, not the inputs. It shows the committed criteria were satisfied by the data supplied; whether the borrower reported that data truthfully is a diligence question, unchanged.
The scorecard that priced this loan is provably the one you committed, so a model cannot be quietly loosened between quarters.
Committed as a bytecode hash. Any change is a new hash, timelocked and announced before it takes effect, with an exit window for capital that disagrees.
The loan file enters the proof as private witness and never becomes a public artefact, so proving the process costs you no confidentiality.
What gets published is the verification result, never the material it was computed from. Witness privacy holds when you run your own prover.
Who this is for
Private credit desks, RWA platforms and lending protocols where the underwriting is genuinely good and genuinely unshowable.
How it works today
Each allocator re-underwrites your process from scratch
Credit-box adherence is asserted in a monthly report
The model can drift with nobody outside noticing
Scaling origination means scaling trust conversations
With POY
One proof per loan, checkable by anyone, no data room
Adherence is a precondition of the loan being written
A model change is announced onchain before it is live
Scaling origination means scaling the book
One proof per loan. No data room, no re-underwriting.