# Independent V2 UX/product + native-feasibility review

**PASS for showing the V2 design/prototype to Owner. No blocking finding remains in this bounded review. V1 remains entirely rejected; V2 DESIGN_APPROVED is PENDING Owner. This is not native SOURCE/BUILD/DEVICE/STORE acceptance.**

The reviewer is independent of the actual `gpt-6-astra/high` author. No new model/provider call, app source modification, credential access, server mutation or publication was performed. Exact reviewed runtime/document hashes are in `independent-review.json`; actual independently executed browser evidence is `independent-browser-result.json`.

## Visual and product assessment

Viewed actual rendered gallery and A390 story, B360 city, C430 guest, A200% story and independently captured C375 city. V2 is a substantial graphic revision: A uses an expansive racket still life and light content progression; B is a distinct energetic typographic/diagonal composition; C uses atmospheric court imagery and dark-to-light city composition. Three coherent story→manual-city→content variants are visible; D/E are honestly labelled concept-only. This satisfies a comparison of real visual concepts rather than recolouring V1. A recommendation is explained, with Owner free to choose. Court imagery is editorial/synthetic context, not a claim about the pictured fictional venue.

The guest path is visible from first story and skips to manual city without auth or OS permission. Stories are skippable, actual useful games follow. Selected second game and Kazan survive wrong/valid code and contextual sign-in; entering the app does not submit a request. Only the later explicit confirmation moves to a synthetic pending result. Return and deep-link examples bypass the introductory story; provider cancellation remains recoverable. Public name is «Игрок», not email/provider name; directory/contact-request defaults and retained opt-outs are described separately from Radar.

## Independent executed browser checks

Owned Chromium1243 via existing Playwright, sandbox enabled, fresh anonymous context on localhost. Eight checks passed, zero JavaScript errors and zero external requests:

- Guest city unknown has honest empty result; manual Kazan preserved without geolocation/auth
- Wrong then valid code preserves selected second game/city and requires fresh confirmation; no account enumeration
- Only explicit new submit transitions to synthetic pending
- Deep link bypasses story; OAuth external simulation and cancel retain recovery context
- Radar consent explicit/unchecked; auth resets consent; OS approximate grant does not enable visibility
- Separate visibility confirmation is required after OS result
- OS denial preserves usable catalog and returning entry bypasses introduction
- Independent compact360/375 × conceptsA/B/C at200%: no horizontal overflow, labelled44px controls and guest CTA reachable by scrolling

The independent compact test covers A/B/C360/375 at200%: no horizontal overflow, labelled44px controls and guest CTA reachable by scrolling. Large text is allowed to scroll vertically; a viewport screenshot is not evidence that the bottom CTA is absent. Existing operator192 geometry +10 interactions and disclosed two CSS-wrap repairs are additional evidence, not substituted for these independent checks. Native screen readers/keyboard and OS permission dialogs remain untested.

## Targeted engineering reconciliation

The existing RN+Expo/ExpoRouter direction remains feasible. Read-only git comparison of `apps/mobile`, API auth, `IOS_ANDROID.md`, `AUTH.md` and safe social defaults from supplied df1d7652 baseline through accepted68 tree-equivalent candidate568fe044 is empty; historical engineering facts are reused honestly without asserting fresh SDK/store availability. There is still no Expo application/native binary, native exchange/refresh, verified app-link setup or APNs/FCM transport. Web email/VK/Yandex methods are not depicted as native-ready; Apple/Google/SMS/password are not invented. iOS4.8/login, store accounts/signing, privacy/deletion and physical devices remain external/Owner gates.

Radar is optional and contextual. Product consent is unchecked, is reconfirmed after auth, and is distinct from OS approximate-location grant and the final visibility enable action. OS and OAuth results are explicitly labelled simulations outside the phone; denied location returns to useful manual catalogue. No background location, automatic Radar, third-party private-message processing or blanket permission request is implied. Exact consent/server eligibility and retention must be bound before implementation; draft UX copy is not new policy.

## Metadata and limits

Resolved V2-META-01: stale README NOT_RUN now explicitly refers to the author-terminal historical stage, followed by actual external/independent evidence. Reviewed final README/index metadata hashes are included; prototype runtime hashes are unchanged from the independent run. Research distinguishes official help pages, historical Apple material and NOT_OBSERVED installed apps; it does not invent installed-app research.

Review covers the new first-launch revision, not a second full native audit. The prototype intentionally does not implement live email/OAuth/auth persistence, real join delivery, native OS permissions, iPhone/Android build or store deployment. V1 readiness is historical technical input, never visual approval. Owner approval, firstnative MVP choice and native/store gates remain pending. Publish only self-contained reviewed public artifacts, excluding private inputs, logs, launchers and user data. No new concept or implementation was requested by this review.