Creator Payout Verifiable Workflow
Platform issues a receipt. Creator re-derives the same payout in their browser. Regulator re-derives from the audit artifact. Three parties, zero trust, one SHA-256.
- 7/7
- Stage hashes match (3 runtimes)
- $0.013
- Float64 drift eliminated
- 0
- Differing cells (204 compared)
- Browser
- Creator verifies in-browser
The scenario
Set the picture
A creator payout at any large platform is the tail of a pipeline: engagement events → fraud adjustment → eligibility gating → revenue-share split → FX conversion → tax withholding → payment. Each step chains arithmetic over tens of thousands of events per creator per month.
Creator payout disputes are structurally high-friction: the creator sees a number, can’t re-derive it, can’t see the audit logs, and has no tool to independently verify. Creator-transparency regulation is converging (EU Platform Work Directive, Brazil, India, US IRS 1099-K).
Cost today
No cryptographic artifact exists that allows a creator to take a month’s input data plus a single receipt and independently confirm the platform’s math is correct.
Float64 implementations produce a $0.013 (12,979 micros) drift vs the integer-exact SolvNum result on the test corpus. The float path happens to agree between two x86_64 hosts — but would diverge on ARM, GPU, or browser.
What changes with SolvNum
Five-stage payout pipeline (rollup, fraud-adjust, eligibility, rev-share, FX+tax) implemented in integer micros. Every stage emits a hash. The final receipt is a SHA-256 over canonical bytes.
Demonstrated across three runtimes: Python 3.12 on Windows, Python 3.12 on Linux (WSL), and JavaScript BigInt on Node 18. All seven hash columns match across all three runtimes. Net payout: 589,331,541 micros ($589.33) identical everywhere.
The JavaScript verifier runs in any browser. The creator doesn’t install anything, doesn’t call customer service, doesn’t take the platform’s word for the number.
Measurable outcome
What we claim — and how it survives review
Each line below maps to a captured number in the demo section. Every number is reproducible from the benchmark suite.
- 7/7 hash columns identical across Windows Python, Linux Python, and Node 18 BigInt.
- Net payout (589,331,541 micros) bit-identical across all three runtimes.
- Float64 implementations drift by 12,979 micros ($0.013) vs integer-exact SolvNum baseline.
- 12 bench rows × 17 non-timing columns = 204 cells compared, 0 differing.
- Tampering detection: single byte flip in events.csv caught as BUNDLE_MISMATCH on any host.
The demo
What was tested. How. What the script printed.
Synthetic creator-month: 100,000 engagement events, 200 content items, 30 days of policy state (including mid-month suspension), 8 countries/currencies, 11 jurisdictions of withholding rules, one revenue-share contract (55%/45%).
Three-sided handshake: platform issues receipt on Windows (→ 38831eec…), creator verifies on Linux (MATCH), regulator re-derives via Node BigInt (PARITY: PASS). Five stress axes: volume scaling, currency complexity, retroactive fraud adjustment, single-byte tampering.
Captured benchmark output
The numbers the script actually printed.
| Artifact | Windows Python | Linux Python | Node 18 BigInt | Match |
|---|---|---|---|---|
| bundle_hash | fe0346… | fe0346… | fe0346… | ✓ |
| stage_1 (rollup) | 9b2da4… | 9b2da4… | 9b2da4… | ✓ |
| stage_3 (eligibility) | a3e622… | a3e622… | a3e622… | ✓ |
| stage_4 (rev-share) | eb1b79… | eb1b79… | eb1b79… | ✓ |
| stage_5 (FX+tax) | 182b09… | 182b09… | 182b09… | ✓ |
| receipt_hash | 38831e… | 38831e… | 38831e… | ✓ |
| net_payout_micros | 589,331,541 | 589,331,541 | 589,331,541 | ✓ |
| Implementation | net_payout_micros | Drift vs SolvNum |
|---|---|---|
| solvnum_backed | 589,331,541 | 0 (reference) |
| float64_naive | 589,344,520 | +12,979 ($0.013) |
| float64_decimal | 589,344,520 | +12,979 ($0.013) |
Float implementations agree between two x86_64 hosts for this corpus but would diverge on ARM/GPU/browser.
Composes with
Where this POC sits in the suite
Evidence pointers
Where the claims live in the repo
These are the files a reviewer should run to re-derive every number on this page.
- tools/solvnum/buyer_pocs/creator_payouts/platform_side.py
- tools/solvnum/buyer_pocs/creator_payouts/creator_side.py
- tools/solvnum/buyer_pocs/creator_payouts/creator_side_wasm/verify.js
- tools/solvnum/buyer_pocs/reports/creator_payouts_xplat_csv_diff.txt
- docs/poc/04_creator_payouts.md
- docs/poc/04_creator_payouts_xplat_evidence.md
Previous · POC 03
Ads Attribution Reconciliation
Next · POC 05
LLM Watermark Detection Reproducibility
Want to see these receipts on your pipeline?
Run the benchmark against your actual decision pipeline.
Two weeks, $25K, fully credited. No production integration, no data leaving your premises. Every claim above traces back to a script you can run locally.