Measured, not asserted
Every product we ship publishes the measurements behind its claims — with a method anyone can rerun, and with the numbers that flatter us least given the same prominence as the ones that do.
What Verilot measures in public
- How Verilot is verified — reproduction from public data, not consensus. Dollar figures only when the balance tie-out matches. Every finding type the engine can raise, with the software step. Machine-readable at /method.json.
- Ordinary-scale case study — five on-chain events, cutoff 2021-12-31, ledger 0.097 ETH matched to the chain. The public example at searcher scale.
- The Phantom Gains Ledger — named public wallets, full method, a balance tie-out on every run, every figure reproducible from public data. Mechanism at extreme scale — our engine, linked vs unlinked, not a measured error in someone else's software.
- The Tool Error Ledger — how tax tools’ figures compare against chain evidence, published only when the underlying scan ties out.
- Every scan result carries its own verification: the tie-out table and a public transaction hash on every flag. When the balance does not tie out, the dollar figures are withheld rather than published with a caveat.
The same discipline, elsewhere in the family
- vet402 (trust scores for AI-agent payments) publishes its measured accuracy — including its own false positives — on its accuracy page, hosted at vet402.com/accuracy.
This page does not wait for numbers that look good. The methodology publishes first, the measurements land as they come, and the empty states say so.