AI news · 2 min read
Codex suggests your next message, and leaves the send button to you
Codex composer predictions proposes a follow-up after a response. The Pro desktop beta needs review before sending; normal message billing still applies.
Announcement date:
Affiliate disclosure: Some links may earn CreatorStack a commission at no extra cost to you.
At a glance
Codex’s October 9 composer predictions beta can suggest a complete next message after a response. Adults on personal Pro can use it in supported desktop local or SSH threads. Tab accepts text; review it before sending. Beta prediction generation uses no credits, but sent messages follow normal billing.
You have a working draft of a page. The useful follow-up might be to inspect its phone layout, rather than ask for another visual change. OpenAI’s October 9, 2026 composer predictions beta proposes a next message from the current Codex conversation. The suggestion may help you continue, but you choose the instruction. Dated release entry.
Accept text, then decide what to ask
After Codex responds, a complete message may appear in the composer. Press Tab to accept it, edit it if needed and send separately. It does not complete words as you type, and it may omit a suggestion. How predictions work.
For a creator building a site or small utility, the important check is whether the proposed next step serves the actual job. A request to polish colours is premature if a checkout or download link still fails. Our example is illustrative: replace a vague “make it better” suggestion with “check the mobile layout and the download link, then show me the result.” We have not tested prediction quality.
The beta has a specific audience
Access is for personal ChatGPT Pro users aged 18 or older, in supported regions, using the latest Codex desktop app. Supported threads are local or SSH with GPT-6 Astra or GPT-6.1 Sol. It is enabled by default; General → Composer → Show predictions turns it off. Generating predictions during beta consumes no Codex allowance or credits, while messages you send follow normal usage and billing. Eligibility, controls and billing.
Try next: review one suggested follow-up against your unfinished work before accepting it. If your immediate question is about the model rather than the message, see our GPT-6.1 Sol coverage.
Common questions
Questions people ask.
- Will accepting a prediction continue the task automatically?
- No. Tab adds the suggested text to the composer. You can edit it and decide whether to send it.
- Why is there no suggestion after a response?
- The feature may skip a prediction. Check the latest desktop app, personal Pro eligibility, supported model and local or SSH thread; you can keep typing without waiting.
Sources & freshness
This article was refreshed on 10 Oct 2026 using the sources below. A source check verifies published information, not output quality or checkout availability in every region.
- help.openai.com/en/articles/6825453-chatgpt-release-notes
- help.openai.com/en/articles/20001601-composer-predictions-in-codex
What changed: New reporting on a desktop next-message beta, separating suggested text, user-controlled sending and ordinary Codex usage.
Found a correction? Let us know.