Article

OpenAI 28 Day Product Sprint Day19: Tibo is reheating old news again,September bringing the cache optimization back to the table

Beating AI flash,OpenAI’s 28-Day Product Sprint has reached Day 19, and Tibo has served up another familiar improvement today: switching reasoning intensity between Codex and ChatGPT Work in a conversation will no longer invalidate the context cache.

But this technology was released as early as the beginning of September alongside GPT-6 Astra, when the Responses API already supported it. Codex was subsequently adapted as well, but it was initially disabled by default. Now, more than a month later, Tibo is presenting it again as a Day 19 update.

Other updates include: 1. dot voice-call connections are faster. 2. Codex CLI added reusable prompt presets, enabling model and permission settings to be adjusted directly in the terminal, and also supports enterprise MCP login. 3. Fixed the failure to resume long CLI sessions and added support for Windows enterprise proxy authentication.

Tibo also disclosed that since DevDay at the end of September, usage of Codex CLI has doubled.

Original link https://m.theblockbeats.info/flash/373490