[Rinat S (rsolmano)]

Problem

The README opened with a wall of signing/notarization caveats, then spent half its length on contributor material that CONTRIBUTING.md already owned. Someone landing on the GitHub page learned how the repo is laid out before learning what ThinkRail does or seeing it. #514 raised this; this PR takes the same goal from the maintainers' side, with the current product vision and screenshots.

Approach

Make the README the user-facing front page and let CONTRIBUTING.md own developer material:

• Lead with the product vision, a workbench screenshot, and a short What you get list derived from the Capabilities in goal-and-requirements.md (so every claim traces to the product spec). The worktree claim admits the Default workspace rather than promising main stays clean.
• Install keeps every fact from the old section but is reordered for a downloader: per-platform desktop assets first, CLI one-liners, then channel/pin/update/uninstall details in a collapsible block, then a Platform notes table. Signing qualifications are kept verbatim in substance.
• Analytics & privacy mirrors the in-app consent copy (AnalyticsPreferences.tsx) and nothing more; the detailed event catalogue stays in the analytics spec.
• Developer material (launcher commands, Electrobun packaging/smoke, e2e:binary/e2e:desktop, the source-update caveats) merges into CONTRIBUTING.md without duplicating what it already had; the stale repo-layout block is replaced by a pointer to architecture.md, which owns the complete list.
• Screenshots are real captures (Claude Sonnet 4.5 via JetBrains AI, on a disposable clone of this repo under an isolated data dir), 1440×900 2x, shipped as dark/light pairs via <picture> so they match the reader's GitHub theme.
• "mobile-first" is dropped: the shell spec describes the mobile shell as future.
• AGENTS.md gains one Handoff-hygiene bullet so README claims and screenshots are reviewed when a user-visible capability, install, or analytics behavior changes.

Every decision was reviewed in three rounds by a second model (gpt-6-astra); its blocking findings (over-claiming "no personal data", Intel "build from source" in the wrong table column) are applied.

Changes

• README.md rewritten; .github/readme-assets/{workbench,specs}-{dark,light}.png added (150–190 KB each, no metadata).
• CONTRIBUTING.md: launchers, packaging/smoke, e2e:binary/e2e:desktop, unstamped-build updater caveat, smoke≠notarization note, repository-layout pointer.
• AGENTS.md: README-currency rule under Handoff hygiene.
• No code or spec changes.

Screenshots

Not applicable — no frontend changes. The README itself embeds the new screenshots; view this branch's README on GitHub in light and dark mode to see both pairs.

Related issues

Supersedes #514.

Checklist

• Fast gates pass: bun run lint, bun run typecheck, bun run test — lint and typecheck ran in the pre-commit hook on the final tree (green); bun run test not run: documentation-only change, no code touched
• E2E suite passes for app-affecting changes (bun run e2e, or bun run e2e:full when touching agent behavior) — not applicable, documentation only
• Before/after screenshots are included for frontend changes, or marked not applicable
• Relevant SPEC.md / top-level specs updated to reflect any boundary, contract, or behavior change — none changed; AGENTS.md updated for the new README rule
• I have read the Contributing guide and agree to the Code of Conduct

[jetbrains-air]

Requesting changes — please address the blocking inline finding.

[jetbrains-air]

Requesting changes — please address the blocking inline finding.

[jetbrains-air]

Requesting changes — please address the blocking inline finding.

[jetbrains-air]

Requesting changes — please address the blocking inline findings.

[jetbrains-air]

Approved — ready to merge.

[jetbrains-air]

Approved — ready to merge.