INFRASTRUCTURE / INDEPENDENT VERIFICATION

AI work. Independently verified.

An AI agent says the code is fixed. ResultBond checks the exact work against qualified hidden tests — then signs a Proof Receipt anyone can inspect.

A verifiable record of what was tested, what passed, and what still needs a human.

Signed evidenceEd25519 Proof Receipts
Clear audit trailExact work and test policy
Honest verdictsPASS · FAIL · INCONCLUSIVE
ResultBond PROOF RECEIPT / ARCHIVE
ENTRY NO.RB-TESTNET-00103 OCT 2026
PUBLIC TESTNET EXAMPLE · DEMO SIGNING KEY
RecordEvidenceStatus
AI code submissionDigest bound✓
Acceptance testsChecks passed✓
Policy decisionALL_CHECKS_PASS✓
OVERALL RESULTPASS
RECEIPT IDreceipt:ea9cf3d6…
POLICYrb.hidden-bugfix.v0.2
SIGNATUREEd25519 · demo key
A SIGNED RESULT — NOT A PROMISE THAT ALL SOFTWARE IS BUG-FREE.

01 / WORKFLOW

Built for verification.

Four precise steps from an AI-generated patch to a receipt you can independently check.

  1. 01

    Agree the job

    Specify the code, acceptance rules and hidden-test policy.

  2. 02

    Freeze the work

    Bind the submitted patch to an immutable digest.

  3. 03

    Run the checks

    Evaluate in a restricted sandbox. Record the evidence.

  4. 04

    Sign the receipt

    Receive PASS, FAIL or INCONCLUSIVE — with reasons.

02 / TRACEABLE EVIDENCE

Every check leaves a trace.

ResultBond keeps a tamper-evident record: what was agreed, which code was examined and what the checks actually concluded.

STEPRECORDED EVIDENCETYPE
01Signed acceptance manifestTERMS
02Qualified hidden testsCHECK
03Sandbox run evidenceRUN
04Ed25519 Proof ReceiptSEAL
Read the engineering specification →

WHAT MAKES IT DIFFERENT

Proof, not promises.

A code review is an opinion. A Proof Receipt records a specific evaluation and can be verified separately from the agent that wrote the code.

  • Checks bound to the exact submitted code
  • Signed receipts with independent verification
  • Hidden tests qualified against known-good fixes
  • Unclear evidence recorded as INCONCLUSIVE

No verification can guarantee that software is universally correct. We publish our known misses.

Read our limitations and misses →

03 / BUILT FOR DEVELOPERS

A trust layer for AI work.

Independent evidence for the teams who build agents, use them, and operate the platforms between them.

01 / BUILD

AI agent developers

Show how an agent performs on checks it never saw, not just self-reported benchmarks.

Explore real cases →
02 / BUY

Engineering teams

Evaluate bug fixes against agreed criteria before deciding whether to accept the work.

Request a shadow pilot →
03 / RUN

Agent marketplaces

Use signed outcomes as inputs for auditable acceptance and future settlement workflows.

See testnet experiments →

RESULTBOND · CHROME EXTENSION

Proof travels farther when it is easy to share.

ResultBond for Chrome finds signed Proof Receipts in GitHub pull requests and checks signatures locally. One-click PR evaluation is still in development.

04 / FAQs

Questions, answered.

Clear boundaries matter more than broad promises.

Is a PASS a guarantee that the code is bug-free?

No. PASS means the specified checks passed under the recorded policy. It is not a guarantee against every defect, security issue or future regression.

Can I verify a receipt without installing anything?

Yes. Open the browser verifier and paste a receipt. The signature check runs locally in your browser.

Are live payments already available?

No. On-chain settlement is currently a testnet demonstration using test tokens. Real-money deployment requires further security and legal review. Shadow pilots move no money.

Which code is supported?

Python coding work is supported in the current pilot. ResultBond's broader integrations are being developed.

READY FOR YOUR FIRST PROOF RECEIPT?

Make AI work accountable.

Begin with a free shadow pilot — no payments or crypto required.
Request a pilot