# Fractal Model Desk v2.1.0 — verification

Verified release dated 8 September 2026. This page separates the producer's native Pine evidence from the source and packaging checks repeated during the web import.

## What was checked

- **89 of 89 package payloads** match the supplied manifest, and the complete ZIP matches its supplied SHA-256. The copied indicator and strategy strings reproduce their Pine downloads byte for byte. Rebuilding the supplied package in isolation reproduces all 89 payloads and the manifest exactly.
- **18 of 18 local source, deterministic build and package tests** were independently rerun. These are not native Pine execution.
- **228 of 228 native assertions across 50 fixtures** are supported by the producer's bundled runtime captures: [raw core: 91/91](v2.1/docs/evidence/v2.1/native-core-qa.json), [delivery/context: 86/86](v2.1/docs/evidence/v2.1/native-context-qa.json), and [independent IC: 51/51](v2.1/docs/evidence/v2.1/native-ic-qa.json). Native Pine was not rerun during this web import.
- The supplied [compiler report](v2.1/docs/evidence/v2.1/compile-results.json) records zero errors and warnings for both final scripts and all three probes. Actual EURUSD and gold captures are included; counts and execution smoke checks do not establish trading performance.
- Comparing the bundled before/after drawing captures again produced **142 identical non-live lines, 140 labels and 18 boxes**. This repeats the comparison, not the original TradingView remove/re-add operation. The explicitly forming preview and per-instance IDs are excluded.

## Scope and limits

Delivery review uses disclosed V-reversal, body/ATR, break-distance and overlap heuristics. The stricter IC review covers a previously known parent-FVG subset. Broader POI and bias still require manual assessment. A local swing sweep does not prove inducement intent, and an automated review is not full model validation or a buy/sell instruction.

Native 5-minute Bar Replay was unavailable; this release does not claim a native Replay pass. Prefix fixtures and same-history reconstruction provide narrower evidence. Changed history, feeds, timeframes or settings can recalculate results, and retained history has a capacity limit.

No protected-indicator parity, trading edge, or future return is established. Results measured for v1.1 or v2.0 do not validate this release.

## Source and documentation

- [Complete original source ZIP](fractal-model-desk-v2.1.0.zip) · [ZIP checksum](fractal-model-desk-v2.1.0.zip.sha256) · [89-payload manifest](v2.1/MANIFEST.json)
- [Setup Atlas source](fractal-model-desk.pine) · [optional Strategy Lab](fractal-model-desk-strategy.pine) · [MIT license](LICENSE)
- [Thai beginner guide](v2.1/docs/start-here-th.md) · [release notes](v2.1/docs/release-notes-v2.1.md)
- [Full supplied verification and linked evidence](v2.1/docs/verification-v2.1.md)
- [Primary-source research](v2.1/docs/research-v2.1-primary.md) · [actual open-source research](v2.1/docs/research-v2.1-open-source.md)
- [Historical v2.0 verification](v2-verification.md)