[RCmerci]

Journal only used Jane Street Core for two loading-delay measurements and a five-second notice lifetime. Measure elapsed time with Mtime 2.1.0 counters, keep the notice and existing sleep scheduling in seconds, and remove Core plus the audited unused direct Base dependency.

The Dune/opam declarations now use mtime.clock / mtime. Each lock drops Core and 52 packages exclusive to its closure; retained package versions, Git pins, shared sexplib dependencies and compiler base-* packages are unchanged. Startup/detail owner checks and the 100ms loading boundary retain their existing behavior, including cancellation and foreground resume.

Validation applies to Core-only implementation 45f845c on main 87d0ccc; subsequent commits change only the decision Markdown, with all source/test/dependency bytes unchanged:

• Targeted startup/routes/source-boundary checks, dune build @all app/native_embed.exe.o, and full forced tests passed: 21 Alcotest suites / 907 cases, plus 115 named running checks reported separately.
• Exact-version production/build dependency traversal has no Core or Base. Native link inputs decreased from 136 to 85 .cmxa archives; both built Simulator and iPhoneOS binaries have zero defined/undefined Core or Base symbols and retain Mtime.
• Same-path, same Release/compiler/SDK/Swift/LUI/signing comparison measured signed executable and App-file bytes:

Target	Executable bytes before → after	App file bytes before → after
ARM64 iOS Simulator	38,023,216 → 29,622,656	39,884,448 → 31,483,888
ARM64 iPhoneOS	38,089,744 → 29,672,768	39,975,900 → 31,558,924

These are actual file sizes for the combined Core/Base cleanup, not compressed download sizes or independent package attribution.

Limits: production Simulator sign-in was exercised, but Core-only synthetic detail/foreground UI acceptance remains incomplete because startup/fixture issues also reproduced in the baseline. A separately developed local-restoration repair, its 915-case run and its successful offline UI acceptance are excluded from this PR. Both Apple builds use the explicit experimental host-restamp path with cross_abi_validated=false: target C/Swift, link/platform/provider and signature checks passed, but a complete OCaml iPhoneOS cross ABI and physical-device runtime are unverified; no iPhone installation was performed. Decision-document validation passes for this change; the full document check still reports three unchanged historical documents missing sections.

GitHub Journal CI run 39032923420 passed on exact head 7703c9a9ab55b8d3a4298cdba090a9508bcfb32e, including all-target/native-object build, full regressions, helper checks and final cache verification. Its uploaded installed-package list also contains no Core or Base (compiler base-* packages are retained).

A fresh independent read-only journal-code-review session completed on published head 7703c9a9ab55b8d3a4298cdba090a9508bcfb32e with no evidence-backed findings. It independently confirmed the GitHub-exported diff matches the local exact range and excludes the separate restoration repair. No merge is requested.

[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 2026-10-17T12:34:46.489649Z	7703c9a	PR opened

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.