Tibo's Day-19 ship lands: faster dot calls and reasoning changes that no longer blow the cache

#What happened

Thibault Sottiaux (@thsottiaux), OpenAI's Codex product lead, posted on X on Oct 23, 2026 at 4:12pm PDT (23:12 UTC): "Day 19/ Small updates. Calls connect much faster for dots, changing reasoning in codex & work mid conversation does not invalidate cache anymore and a we've also updated the codex cli." (typo and all — the post drew around 5,000 likes at last check).

This is Day 19 of his 28-day challenge, announced Oct 4: each day he either ships a clear improvement or triggers a full reset, with the run going to about Nov 1. And it fulfills the "small ship" he promised alongside the Day-18 global Codex reset, which landed this morning at 10:40am PDT (already covered on the site).

#What changed

The two improvements, in his words: (1) "Calls connect much faster for dots" — Dots are OpenAI's always-on GPT-6 Astra agents inside ChatGPT; (2) "changing reasoning in codex & work mid conversation does not invalidate cache anymore." He also notes "a we've also updated the codex cli" — Codex CLI 0.164.0 stable shipped Oct 22 (reusable prompt presets, inline model/permission settings, image preview, enterprise MCP sign-in), already covered separately.

#Why it matters

Cache preservation means reasoning changes stop burning extra quota: cache misses burn allowance faster, so this one directly protects your budget. And faster dot connects reduce the friction in Dots calling you back.

Verified 2026-10-23.