Research explanation and public-copy integration

Date: 2026-09-27

Scope

Complete the next editorial step: connect the founder's learning motivation with concrete research questions and the current deterministic planner. The maintainer also requested a pull request for the accumulated, previously local public-copy work.

Added paired DE/EN research pages covering request interpretation, preference research, search guidance, personal patterns and possible later applications. They explain the reserved from-scratch learning path without implementing it. Current ranking observations, Phase-A evaluation wire types and ML metadata/interfaces remain distinct from training, collection, replay and learned planning. Linked the research explanation from the existing homepage, roadmap prose, introductions, architecture references and language indexes.

The first-person origin uses the maintainer-supplied story. The paper's year and three authors were checked against the publisher's page for Learning representations by back-propagating errors. Private handoff files were not copied into the repository.

Integration

Kept the canonical workspace and preserved the unrelated .gitignore change and root test output. Created codex/public-explanation from origin/main at 2a98cf8, after the earlier UI and evaluation PRs had merged. Applied the saved copy changes there, retaining main's layouts, animation and newer components. Moved the comparator explanation to its current component and updated existing button-text assertions. No CSS, planner behavior, schema, versions or release publication changed.

The PR combines the already completed introduction, CPIR and architecture explanations with this research conclusion as one public explanation change. The earlier dated logs describe the checks on their original branch; this record qualifies the integrated result. Full interactive-site localization, a backpropagation simulator and further UI work are not part of this PR.

Verification

• Repository checks passed, including 22 version-authority tests, links, DE/EN metadata and dependency boundaries.
• The regular site check passed with zero errors and warnings, including Rust data generation.
• The production build passed with the GitHub Pages base path /Nexus-Cerebri.
• Additional Chromium checks covered the roadmap and eight DE/EN explanation routes at 1440, 390 and 320 pixels: HTTP 200, correct article language and no document overflow. Axe found no serious or critical violations at 390 pixels. Research language links and 200% text checks passed. A source link in the introduction's input table restored keyboard access to its horizontal content on narrow screens.
• All 68 non-visual Playwright tests passed on the integrated site. The first run exposed two old button labels in an upstream continuity test; only those text selectors changed.
• Reviewed all 28 Windows visual captures, with detailed mobile crops for the changed explanation, comparison, measurement and founder passages. Updated the Windows hashes and reran the unmodified visual gates: all 28 passed.
• Checked semantic DE/EN parity, equal inline technical terms and unchanged component style blocks. git diff --check and the staged whitespace check passed.
• The initial PR CI run 36310686211 on c4bd764 passed 14 of 15 gates, including the Linux browser suite. Only the visual hash gate differed from the pre-copy references. Reviewed its 28 Linux captures and detailed mobile crops; every initial/retry pair had identical hashes. Updated only the Linux reference hashes from that artifact. The follow-up revision must pass the actual remote gates before this PR is marked ready for review.
• No release, tag, merge or deployment is authorized by this publication of code for review.

This is implementation evidence, not independent Architecture, Math or Security qualification.