Day 2 of Tibo’s 28-day Codex improvement plan is here, and the main update is that Auto-review is now free for all users signed in through a ChatGPT account. For people running longer agentic tasks in Codex-like workflows, the immediate effect is less manual clicking through approvals under the default sandbox behavior.
Tibo said earlier that over the next 28 days, the team would either ship one clear improvement relevant to most Codex/Work users each day, or carry out a full reset. The free rollout of Auto-review is the Day 2 move in that daily improvement push.
What the Day 2 change does
Auto-review adds a second review layer on top of the main agent. When it is enabled, a second agent reviews all actions taken by the primary agent while long tasks are running. According to Tibo, its only goal is to prevent high-risk actions and protect against unwanted moves that do not match the user’s original intent.
That is meant to fix a known pain point in the default sandbox setup: if every action requires manual approval, users can quickly hit decision fatigue. Unless they spend a lot of time writing specific rules upfront, long tasks can become a long chain of approval prompts. Auto-review is presented as a way to reduce that burden.
How users turn it on, and what it changes
The setting is available at settings > permissions > auto-review. The post does not describe extra eligibility requirements beyond being signed in through a ChatGPT account.
There are two practical takeaways for users. First, long tasks can keep moving with less manual supervision, because the second agent handles many reviews automatically. Second, Tibo says the feature is now free and does not draw usage from a user’s plan.
A diagram attached to the post shows the flow with example numbers: out of 10,000 actions, 9,280 stay inside the sandbox and 720 go through Auto-review. Of those reviewed actions, 713 are approved and continue, while 7 are denied: 4 continue via a safer alternative, and 3 stop and ask the user.
Before this Day 2 post, the same thread mentioned a Day 1 speed optimization: default speed across GPT-6 Astra and GPT-6.1 Sol was improved by roughly 50% across subscription products and partner products using Sign in with ChatGPT. But the center of Day 2 remains the free availability of Auto-review.
One point worth keeping precise: Tibo’s phrase “full reset” is not explained in the available posts. Since the post does not specify what would be reset, that wording is left as Tibo’s own statement rather than a confirmed product detail.