Skip to main content

Solutions

Reporting, smart contracts, and on-chain security.

A figure that is nearly right is not right. A contract that usually keeps its balance is not safe to hold funds. Underneath, these are the same engineering problem: a rule that has to hold every time, with evidence a person can read.

The method behind all of it

  1. 01

    Write the rule down.

    We start from what has to be true, not from the code or the spreadsheet. What must always hold, what must never happen, what a total has to equal.

  2. 02

    Check it mechanically.

    A machine checks the rule, not a reviewer's judgement. In Inference the rule is checked against every possible input. In Haruno every figure is traced to the transactions behind it.

  3. 03

    Report the specifics, not a verdict.

    A green tick tells you nothing you can act on. We report the input that breaks a rule, the transactions behind a movement, or the piece of data that is missing.

  4. 04

    Keep it checked as things change.

    Correctness is not a milestone. The checks run again when the code changes and when new data arrives, not once in a report.

What we will not claim

Inference does not cover floating-point arithmetic or machine-learning models, and we will say so in the first conversation rather than the third. Haruno is not an audit and does not make the judgment calls for you. We do not issue security certificates or stamps of approval, we do not manage money or give investment advice, and we do not take on work we cannot staff properly.

Start a conversation

A figure that is nearly right is not right. A contract that usually keeps its balance is not safe to hold funds. Underneath, these are the same engineering problem: a rule that has to hold every time, with evidence a person can read.