# Validation — Compound Interest Accrual

Reviewed25October2026. Private package; no public deployment, encoded/narrated video, accepted journal, authenticated source or live SQL write is claimed.

| Gate | Result | Evidence |
|---|---|---|
| Before-code oracle |15 independent polynomial/weighted rational cases | [independent algebra](../datasets/independent-algebra.json), [exact requests](../datasets/independent-cases.json) |
| Python |3 methods pass, including77 fixture subcases and two-capitalization continuation tests | test_compound.py (canonical test source) |
| Whole-output parity |77 Python/TypeScript responses equal, including all exact state/trace/proposal fields and typed null rejections | [parity-results.json](parity-results.json) |
| Large exact output |4096 digit opening value grows beyond 4300 output digits; closed-form expectation agrees; no global integer settings changed | Fixture V06 and test exact polynomial |
| Calendar/carry |Complete35 days; leap month-end; no forced Nov4 cap; rate causality; carry±half; zero legs and amount-limit hypothetical state | Shared corpus plus continuation properties |
| Browser |423 states:4 success scenarios×35 days×3widths plus missing-day rejection at 3widths | [browser results](browser-qa/results.json) |
| Controls/layout |390/768/1440; no document overflow; buttons/selects≥44 px; reset, rate changes, Back/Step/Play/Pause, keyboard and reduced-motion one-step pass | visual-qa.mjs (canonical runner) |
| Console |0 warnings/errors | [browser results](browser-qa/results.json) |
| Assets |Both staticSVGs and Mermaid rendered; images inspected | [phone](browser-qa/lab-390.png), [tablet](browser-qa/lab-768.png), [desktop](browser-qa/lab-1440.png), [pools](browser-qa/pool-flow.png), [calendar](browser-qa/calendar-comparison.png), [workflow](browser-qa/mermaid.png) |
| Skill validators |Package0errors/0warnings; light playground1pass/0fail | Local validation commands |

Run from the topic folder with repository Node/TypeScript and workspace Playwright runtime installed:

```text
python -B -m unittest discover -s tests -p test_compound.py
python -B tests/run_parity.py
python tests/build_lab.py
node tests/visual-qa.mjs
python examples/demo.py
```

The runner compiles TS in strict ES2022/NodeNext mode to ignored `tests/.tmp-compiled`. Lab generation embeds that actual compiled financial kernel and complete canonical35-day baseline; rate/scenario controls create labeled synthetic comparison branches. All data is author-created. Approximate lab numbers are truncated display labels beside exact retained fractions; they are not arithmetic inputs or settlement rules.

The blocked-credit case preserves hypothetical scheduled Q/state but has no posting legs. That state must not be mistaken for accepted customer credit. Source authentication, payable funding, account authorization, prior accrual reconciliation and Stage 1 acceptance remain host obligations. The continuation tests establish consistency inside the explicit resource/schedule profile, not unbounded state growth or actual ledger continuity.

The independent oracle used closed-form cash-flow algebra before the engine existed. Full35-day daily delayed interest≈442.515457121; month-end32240/73; daily Oct31credit401 changes subsequent principal and exact earned≈442.515532429. Minor differences from credit quantization are explicitly modeled; no product-return claim is inferred.

Root separately owns visitor sync, production build, route/asset HTTP and embedded browser gates. Local private readiness does not assert those public-surface checks have run. The article links the publishable copied evidence under examples rather than research paths.