Codex's 28 days: every improvement and reset so far

On 4 October, Tibo (@thsottiaux), who leads Codex at OpenAI, started a public challenge for Codex and ChatGPT Work users. For 28 days, every day ends with either a real improvement or a full usage reset. We'll keep this tracker updated. Everything below links to his original posts.

The warm-up: a reset before the series

• Tibo announced a global reset for all paid ChatGPT accounts and apologised for GPT-6.1 Sol's slow start after a huge load spike in its first two days. He then confirmed the reset had propagated.
• Some Pro 500 users didn't get it at first. On 3 October Tibo said it was "All fixed."

Day by day

• Day 1 (5 Oct), speed: the default speed for GPT-6 Astra and GPT-6.1 Sol on the subscription is ~50% faster across OpenAI's products and Sign in with ChatGPT partners such as OpenCode, Pi, Amp and Devin. Post
• Day 2 (6 Oct), four ships: Auto-review is free for ChatGPT-signed-in users and doesn't draw on your plan; a simplified API for builders; integrated meeting notes; and a Decisions API. Auto-review · Roundup
• Day 2, then a reset anyway: after a community vote, Tibo said "the community demands a reset" and processed one. Post
• Day 3 (7 Oct), GPT-6 plus a banked reset: Tibo called GPT-6 in ChatGPT "the big one", noted a new high of 40M active users across Codex and ChatGPT Work, and loaded a banked reset into paid accounts. He confirmed it had landed. Post
• Day 4 (8 Oct): we couldn't find a post labelled Day 4 on Tibo's public timeline, so we've left it blank rather than guess.
• Day 5 (9 Oct), composer predictions: the desktop app can suggest your next message. It's included in Pro plans without using your quota, and OpenAI Developers calls it a beta. Post
• Day 6 (10 Oct), reset: "Day 6/ Reset", then "Global reset by EOD". Tibo later confirmed it was all propagated. Day 6 · Announce · Done
• Day 7 (11 Oct), 500k context: the default context for GPT-6.1 Sol in Codex and Work goes from 272k to 500k (~80% more). You can still set it anywhere up to the 1M maximum, and there are no long-context fees when usage comes from your subscription. Post · Fees

Side quest: the Pro limits poll

On 11 October Tibo asked Pro users to choose: "5h limits, 2X week usage" or "no 5h limits, same usage". With 240,643 votes, the final result was 37.8% for no 5-hour limits, 33.2% for double weekly usage, and 29% for just seeing the results. Tibo said the result confused him, because he expected more people across Pro 100/200/500 to pick 2X usage, even with 5-hour limits coming back. Nothing has changed yet; it was only a poll.