REPLAY / SIMULATION

Frozen protocol fixtures. No live transaction, AI attestation or physical verification is claimed.

Mode & demo role
MENDMaintenance with memory
ClosedREPLAY / SIMULATION · FIXTURE FALLBACK2 reports

Riverside Park bench — split seat and loose armrest

Two nearby reports merged into one repair. A sealed contractor market selected the strongest total value, a human inspector accepted completion, and the 10% durability holdback released after a second inspection.

Riverside Park, bench near east gate WO-BENCH-001-0001 contractor-b….testnet

Fixture fallback. The live API has no case record for this id; the frozen replay fixture is shown instead. No live transaction, attestation or physical verification is claimed.

Current state · plain language

The durability inspection passed, the holdback released once and the work order is financially immutable.

Next: No further payment is possible. The history remains in the asset passport.

Updated 10 Nov 2026 · 13:35 UTC

Money at a glance

100.00 USDC

Funded in mend-usdc.testnet

Refunded
20.00 USDC
Completion paid
72.00 USDC
Holdback held
0.00 USDC
Holdback released
8.00 USDC

Approved scope

What the contractor is allowed to do

Scope, budget, roster and scoring policy were authorized together. AI may propose; it cannot approve.

DAO #42 · APPROVED

Remove the split seat plank and loose armrest. Install a sealed hardwood plank with stainless fixings. Sand exposed edges, treat end grain, re-fit armrest, remove waste and submit six-item redacted evidence.

Scope hash441a425d…91d092
Scoring policy hash1a8a9858…42357d
Maximum budget
100.00 USDC
Awarded bid
80.00 USDC
Contractor roster
3 approved accounts
Inspector roster
2 approved accounts
DAO authorization is a push path: a Sputnik FunctionCall proposal executes dao_authorize only after approval. MEND does not trust proposal text. This replay reference is not a live proposal lookup.

Full timeline

Every state, actor and evidence reference

No unexplained “processing” state. Each entry names the actor and the decision that moved the case.

16 events
  1. Report

    Report received without a wallet

    Photo and description submitted. EXIF/GPS metadata removed in the browser before upload.

    anonymous reporter · case_7J2mQ9xLk4Pz1VbN

    • Redacted media hashaaaaaaaaaa…aaaaaa
  2. Report

    Second report received

    A nearby reporter describes the same bench damage.

    anonymous reporter · case_3Hs8Kd2Lm9Qw4XcR

    • Redacted media hashbbbbbbbbbb…bbbbbb
  3. Triage

    Duplicate cluster confirmed by a human steward

    Private AI proposed 94% confidence. The steward confirmed the merge; no automatic merge was applied.

    steward.mend.testnet

    • Cluster fixtureCL-0001
    • Human decisionsteward.mend.testnet
  4. Authorization

    Scope and maximum budget approved; DAO proposal attached

    Steward approves the repair scope. The DAO envelope and scoring policy are referenced, not invented by AI.

    steward.mend.testnet + mend-dao.sputnikv2.testnet

    • Scope hash441a425d51…91d092
    • Policy hash1a8a985878…42357d
  5. Funding

    Sponsor funded 100 test USDC

    Exact-token, exact-amount funding into the work-order escrow. Test value only.

    sponsor.mend.testnet

    • Funding tokenmend-usdc.testnet
  6. Sealed bids

    Sealed bidding opened

    Three rostered contractors may commit. Competing values are invisible until the reveal phase.

    steward.mend.testnet

  7. Sealed bids

    Three sealed commitments recorded

    Commitment hashes are public; price, duration and warranty remain sealed.

    3 rostered contractors

    • BID-A-BENCH-0011e1b013855…2d67a0
    • BID-B-BENCH-001fcb0780b0d…79d968
    • BID-C-BENCH-001008850641e…dbfce2
  8. Sealed bids

    Reveals checked against commitments

    Every revealed bid matched its commitment hash, arrived in the reveal window and came from the approved roster.

    Hub contract

  9. Deterministic score

    Deterministic score produced

    A hash-pinned WASM policy ranks total value. No LLM chooses the winner.

    OutLayer execution (replay fixture)

    • WASM hash790f3eb4e0…a3ef42
    • Input hashf44bd67a69…35e947
    • Output hash792aba5f6d…71304d
  10. Award

    Steward confirmed the rank-1 eligible bid

    contractor-b.mend.testnet awarded at 80 USDC. The unallocated 20 USDC returns to the sponsor.

    steward.mend.testnet

    • Human award confirmationsteward.mend.testnet
  11. Work

    Contractor started work

    Only the awarded contractor can start and submit completion for this work order.

    contractor-b.mend.testnet

  12. Evidence

    Redacted completion evidence submitted

    Six checklist items expected. Evidence is stored privately and referenced on-chain by redacted hash only.

    contractor-b.mend.testnet

    • Completion evidence629d602928…40fb88
  13. Inspection

    Independent human inspector accepted completion

    The inspector is not the contractor. Acceptance releases exactly 90% and starts the durability window.

    inspector.mend.testnet

    • Human acceptanceinspector.mend.testnet
    • Notes hashcaef7d9ff1…44a9f2
  14. Payment

    72 USDC completion payment recorded

    90% of the 80 USDC award. The remaining 8 USDC is held for durability.

    Escrow

    • No live transaction hash in replay
  15. Inspection

    Durability inspection passed; 8 USDC holdback released

    Same work order. Same asset history. The durability timestamp was fixed before the repair began.

    inspector.mend.testnet

    • Human durability decisioninspector.mend.testnet
    • Physical observationSeat and armrest stable after loading check
  16. Payment

    8 USDC holdback released; work order CLOSED

    Total: 20 refunded + 72 completion + 8 holdback = 100 funded. Released exactly once.

    Escrow

    • No live transaction hash in replay

Payment rail

Every cent accounted for

Ledger balanced · 100.00 = 100.00

Sponsor budget

100.00 USDC

Funded into escrow

Unawarded refund

20.00 USDC

Returned to sponsor

Completion payment

72.00 USDC

90% after independent acceptance

Durability holdback

8.00 USDC

10% released after durability pass

Payment history

mend-usdc.testnet
Payments recorded for WO-BENCH-001-0001
KindAmountRecipientStatusRecorded
refund20.00 USDCsponsor.mend.testnetSUCCESS10 Nov 2026 · 09:25 UTC
completion72.00 USDCcontractor-b.mend.testnetSUCCESS10 Nov 2026 · 11:30 UTC
holdback8.00 USDCcontractor-b.mend.testnetSUCCESS10 Nov 2026 · 13:35 UTC

A completion payment and holdback are calculated exactly once. A closed work order cannot be paid again. A warranty reopening creates no new escrow.

Open sponsor & Intents view

No payment above carries a real transaction hash in this replay fixture, so no explorer link is shown.

Work order stateClosed

Evidence packages

Private media, public hashes

On-chain state references redacted hashes only. Raw media and precise location stay off-chain.

1 package

EV-0001 · completion evidence

contractor-b.mend.testnet · 10 Nov 2026 · 11:00 UTC

Redacted
  • Before photo of the damaged asset
  • Repair or replacement in progress
  • After photo from the same angle
  • Close-up of the repaired joint or fixing
  • Materials or fixings receipt / label
  • Site left clean and safeContractor says sweep completed; no photo supplied.
Evidence hash 1629d6029…40fb88
Evidence hash 2dddddddd…dddddd
Evidence hash 3eeeeeeee…eeeeee
A completeness review can identify missing checklist items. It cannot certify physical quality and it never supplies the human acceptance decision.

Verification ledger

What is verified, what is unavailable, and what is a human claim

REPLAY RECORDING
Verified inference evidence

Cryptographically checked inference, execution or transaction evidence.

Unavailable evidence

Expected evidence that was not available, not bundled or not verified. Never shown as verified.

Human decision

A named human decision. A human signed or explicitly confirmed it; AI did not.

Physical claim

A claim about the physical world. It is not machine-certified and does not imply structural safety.

Private AI triage

NEAR AI Cloud · structured output only

Unavailable evidence

Structured classification, duplicate candidate and scope recommendation were generated in a private inference session. Raw media never reaches the chain.

Replay recording. No live NEAR AI response signature is bundled, so this artifact is shown as unavailable rather than verified.

AI proposes classification, duplicate candidates and scope. It cannot approve scope, award a bid or move money.

OutLayer deterministic execution

Hash-pinned WASM · no LLM ranking

Unavailable evidence

Policy, WASM, input and output hashes are displayed below. The contract would accept only a hash-bound callback from the configured OutLayer account.

Policy hash1a8a9858…42357d
WASM hash790f3eb4…a3ef42
Input hashf44bd67a…35e947
Output hash792aba5f…71304d

Replay recording. The deterministic ranking can be reproduced from the frozen bids, but no live OutLayer callback is claimed.

A live result would be accepted only through the configured OutLayer callback and matching policy, WASM, input and output hashes.

Human decisions

Steward · inspector · DAO

Human decision
  • steward.mend.testnet confirmed duplicate cluster CL-000110 Nov 2026 · 08:20 UTC
  • inspector.mend.testnet approved completion inspection10 Nov 2026 · 11:30 UTC
  • inspector.mend.testnet approved durability inspection10 Nov 2026 · 13:35 UTC
  • mend-dao.sputnikv2.testnet approved proposal #4210 Nov 2026 · 08:30 UTC

Replay recording. Human actors are labelled, but their wallet signatures are not part of this fixture bundle.

A named decision is not the same as a cryptographic signature. Replay fixtures label human actors but do not bundle their wallet signatures.

Physical-world claims

Contractor evidence · inspector observation

Physical claim
  • Seat and armrest stable after loading check10 Nov 2026 · 13:35 UTC · observed claim, not machine-certified

MEND does not certify structural, electrical or gas safety. Physical claims remain the responsibility of the named contractor and human inspector.

Transaction links

No explorer link — no real transaction hash in this replay fixture.

Contract references

Hub
mend-hub-v2.testnet
Token
mend-usdc.testnet
DAO
mend-dao.sputnikv2.testnet

Why hashes matter

Hashes bind the displayed evidence to bytes without publishing private media. The hash alone does not prove a physical repair. It proves that the referenced bytes were the bytes evaluated or recorded.