PROOF

From a claim to evidence

How results are verified

PROOF distinguishes operating products from demonstrations and research from assumptions. Verification connects the task, conditions, result and limitations.

Local checks, CI and results in the published environment are recorded separately.

What makes a result verifiable

Use this sequence to read a case and understand which evidence supports its conclusions.

  1. Task

    What happens
    Describe the current work, required outcome and consequences of an error.
    What becomes clear
    The object of the check is clear.

  2. Conditions

    What happens
    State the data, limitations, author’s role and test environment.
    What becomes clear
    The result can be assessed in its actual context.

  3. Criteria

    What happens
    Define success and acceptable errors. For AI, check answers on concrete examples and assess the human review required.
    What becomes clear
    Useful output is distinguished from a convincing impression.

  4. Result

    What happens
    Show an operating path, demonstration or research artifact with its actual status.
    What becomes clear
    Implemented work is distinct from planned work.

  5. Verification

    What happens
    Compare the result with the criteria; record errors, corrections and repeated checks.
    What becomes clear
    The conclusion has evidence that can be checked.

  6. Limits

    What happens
    State the date, version and checks still pending. A synthetic demo does not establish industrial deployment.
    What becomes clear
    The scope of the evidence is clear.

Ask about evidence or a limitation

If a case lacks information you need to assess it, ask a specific question. The page context is preserved in your enquiry.

Ask a question