fix(web): reduce rendering overhead and repair mobile navigation

[cryptasian]

Initial hero content waited for hidden client reveals, and mobile carousel pagination exposed 8px targets. This repair makes initial content readable from server HTML, keeps safe optional entrances below the viewport, and provides stable 44px pagination with specific service/privacy link names and unchanged consent semantics.

Shared rendering now uses exact lazy skyline vector crops, verified compact font faces with complete original-character fallbacks, native illustration observation and native process accordion transitions. Runtime-only Tailwind sources exclude planning examples. Artwork, copy, responsive layout, phase selection, reduced motion and themes are preserved.

Refs #96, #97. Parent #31. These references do not close their production acceptance gates.

Validation

All 24 cold local HTTP/2 runs pass score >90, LCP <2500ms, FCP <1800ms and CLS <0.1. Every desktop score is 100; every final run has accessibility/SEO/best-practices 100 and no warning/error. All final CLS values are zero.

Mobile route	Scores (three runs)	LCP ms (three runs)	FCP ms (three runs)
homepage	98 / 98 / 98	2254 / 2255 / 2254	1054 / 1055 / 1054
services	98 / 98 / 98	2254 / 2252 / 2254	1054 / 1052 / 1054
about	98 / 98 / 98	2253 / 2254 / 2254	1011 / 1018 / 1011
work	99 / 99 / 98	2120 / 2120 / 2334	903 / 904 / 902

Recorded production requests use HTTP/2; earlier next-start localhost results use HTTP/1.1. The additional controlled baseline/repair comparison uses the same loopback HTTPS/HTTP2 proxy, host and unchanged Lighthouse 13.4.1 default simulated CPU/network settings. All 12 unchanged-main HTTP/2 mobile LCPs fail. Local TLS uses a disposable certificate exception only in owned headless profiles. All 198 formal reports and earlier failures are retained; transport matching is local evidence, not final production acceptance.

• Complete per-run evidence, provenance and limitations — measured source 71ccb6f.
• Exact-source CI passes locked install/audit, lint, explicit TypeScript, 7 contrast + 158 unit/editorial + 8 integration + 1 production-publication tests, deterministic skyline check and optimized build.
• Browser checks retain keyboard phase selection, inert inactive panels, original skyline crops, 44px targets and no overflow at 320/390px. Font generation verifies glyph geometry/hinting/metrics/full coverage and repeated identical output. Earlier navigation/theme/consent/swipe/initial-HTML evidence remains linked.

Latest PR head c343e6d also passes the complete CI workflow; runtime source is unchanged from the measured commit.

Release verification

Implementation is complete and ready for review. #96/#97/#31 remain open for final production measurements, final target/link audits and their genuine field-data gate. Vercel builds the measured source successfully, but anonymous preview requests reach sign-in. Automatic approval review also blocked using the existing browser session without explicit preview-access authorization; that authorization is requested. No protected-preview result, production deployment or field INP success is claimed.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
shruggietech-site	Ready	Preview	Sep 18, 2026 5:05am UTC

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-18T05:10:17.988294Z	c343e6d	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.