Tibo's Day 2 | Free Auto-review, simpler API tiers and a reset
Auto-review stops consuming subscription usage. The day also covers API tiers, the Meetings beta, Decisions API and a later reset confirmation.
On this pageCollapse
The four Day 2 updates address long-running Codex tasks, API developers and meeting workflows. Free Auto-review is the most direct usage change. A reset was confirmed later that day.
Where to enable free Auto-review
The Day 2.1 announcement says Auto-review no longer consumes plan usage for users signed in with a ChatGPT account. A separate agent reviews the main agent’s actions for risk or departures from user intent, reducing repeated approval interruptions during long tasks.
Tibo’s reply points to the permissions menu below the composer: choose Approve for me. The main post also gives Settings → Permissions → Auto-review. Labels can vary by version; compare the official Codex plan guidance.
The roundup says Auto-review previously could account for 2%–10% of plan usage. That describes the author’s estimate of the old review overhead; actual savings depend on the task. Main-task usage and existing workspace permission boundaries continue to apply.
A lower threshold for the top API tier
The Day 2.2 post quoting @OpenAIDevs replaces five paid API usage tiers with Build, Launch and Grow. The cumulative API-payment threshold for the top Grow tier falls from the previous top tier’s $1,000 to $500.
This changes eligibility for higher API rate limits. That announcement does not change model unit prices or ChatGPT subscription limits.
Conditions for Meetings and Decisions API
The ChatGPT Meetings announcement specifies a beta on macOS desktop for Pro and Business, with Enterprise still to come. Search for Meetings in the plugin directory.
The plugin saves personalized meeting notes and next actions to ChatGPT Space, where they can remain private or be shared with a team. Those notes can then support project plans and follow-up work.
The Decisions API is a developer public beta for quickly selecting a model, tool or next action. The official “up to 10× faster” comparison concerns decision-making with GPT-6 Luna through the Responses API, a specific performance claim.
The day’s reset was publicly confirmed
At 03:35 UTC on October 7, 2026, Tibo said the reset had been handled. In Pacific time, it was still October 6, Day 2. The post does not enumerate plans or usage windows; its confirmation stays with this series day rather than becoming a separate brief.
Sources & corrections
Published UTC
