Waiting on skill, memory, then history one at a time? On 0.163.0 Codex runs those read-only lookups in parallel

#What changed

Codex CLI 0.163.0, announced by @CodexReleases on X, lets Codex run independent read-only tools in parallel. When a turn needs several lookups that do not depend on each other, Codex no longer has to finish one before starting the next.

The release notes and merged pull request #52245 call out skill, memory, and conversation-history lookups as examples. Those tools stay read-only; the change is about overlapping work that used to be serialized.

#Who is affected

Anyone on Codex CLI 0.163.0 or later whose turns often pull skills, memory, and conversation history in the same step. If a turn only uses one of those tools, you will not notice a difference. Write tools and anything that must wait on another tool's result still run in order.

#What to do now

• Update to Codex CLI 0.163.0 or later if you are still on an older build.
• On turns that load skills and memory together, watch whether the wait feels shorter; that is the parallel path working.
• Keep treating write actions and dependent tool chains the same way you did before — only independent read-only lookups overlap.
• If you build custom skills or memory hooks, avoid assuming they always run alone; another read-only lookup may be in flight at the same time.

#What is not confirmed

This is based on the official release notes and PR #52245; DotsFeed has not timed before-and-after turns. The notes do not list every read-only tool that can run in parallel, and they do not say whether the Codex app and IDE extension ship the same scheduling.

#Sources

• @CodexReleases on X: https://x.com/CodexReleases/status/2110847076934328501 (Added reply https://x.com/CodexReleases/status/2110847079996096687)
• GitHub release rust-v0.163.0: https://github.com/openai/codex/releases/tag/rust-v0.163.0
• PR #52245: https://github.com/openai/codex/pull/52245