OpenAI’s 28 Days of Shipping: Codex and Work Updates So Far
What Has OpenAI Delivered on Its 28-Day Promise? Follow the Latest Codex and ChatGPT Work Announcements in One Updating Timeline, From Faster Models and New Features to Usage Resets. See What Changed, Who Can Access It, and Which Releases Could Make a Difference to Your Next Task.
Updated regularly · Last updated: October 19, 2026, 09:50 SGT · Newest first · Announcement dates: Pacific Time.
OpenAI's 28-day shipping commitment promises one clear improvement relevant to most Codex or ChatGPT Work users each day, or a full usage-limit reset. It is a forward-looking commitment, not a claim that 28 releases have already happened. This timeline covers Days 1–14, through October 18, 2026, Pacific Time, with original announcements linked below. The OpenAI Developer Community tracker provides additional background.
Latest Updates: Newest First
October 18, 2026 — Day 14 — Edit Earlier Messages in Desktop Threads
Tibo announced that users can now edit earlier messages within a thread in the desktop app, delivering a frequently requested usability improvement. The announcement does not specify operating systems, plan restrictions, or how editing affects subsequent replies. Day 14 announcement.
October 17, 2026 — Day 13 — Another Usage Reset Completed
Tibo announced a Day 13 usage-limit reset and said in the announcement thread that it would arrive by the end of the day in Pacific Time. He later posted a completion confirmation at 05:12 SGT on October 18 (14:12 Pacific Time on October 17). The posts do not specify eligible plans or affected usage windows; check your account’s usage screen for the limits actually refreshed. Day 13 announcement.
October 16, 2026 — Day 12 — More Effective Usage Through Context Caching
Tibo announced efficiency improvements to Codex and Work that increase effective usage by reusing context across subagents, retaining cached context longer, and routing requests to improve cache hits. These changes reuse prior processing so the allowance can support more work; the announcement gives no percentage increase or plan-specific figures. Day 12 announcement.
In the same thread, Tibo also highlighted a Pro upgrade. OpenAI’s announcement confirms faster GPT-6 Astra Pro responses, improved plugin and tool use, and parallel agents whose progress and completed tasks users can review. This upgrade is available to Pro users in the Chat tab.
Tibo also recommended updating the app to try dots on the Home Screen, linking to OpenAI designer Gavin Nelson’s demonstration. The posts do not specify plan or regional availability.
October 15, 2026 — Day 11 — Fewer Model Capacity Interruptions
Tibo said system changes and additional compute have been deployed to significantly reduce the “Selected model is at capacity” message. If the message still appears, the request should eventually resume automatically instead of remaining blocked. He asked users to report continuing cases. The announcement does not specify affected plans or models, or give a guaranteed recovery time. Day 11 announcement.
October 14, 2026 — Day 10 — Usage Reset Completed
Tibo announced a usage reset for Day 10. He initially said it would arrive within two hours, then confirmed completion at 14:28 SGT on October 15 (23:28 Pacific Time on October 14). The posts do not specify plan eligibility or which usage windows were refreshed; check your account’s usage screen for the reset actually applied. Day 10 announcement · Rollout timing.
October 13, 2026 — Day 9 — Less Disk Space for Desktop Apps
Tibo announced that storage optimizations have rolled out for the ChatGPT and Codex desktop apps. OpenAI expects them to reduce the amount of state stored on disk by about 60%, addressing feedback from heavy users running out of disk space. This is an expected reduction in stored application state, rather than a measured saving for every installation. The announcement does not specify plan restrictions. Day 9 announcement.
October 12, 2026 — Day 8 — Another Global Usage Reset
According to the OpenAI Developer Community tracker, Tibo announced another global usage-limit reset, scheduled to arrive by the end of the day in Pacific Time. In a follow-up post, Tibo confirmed that the reset was complete. Check your account’s usage screen for the limits actually refreshed. Day 8 announcement.
October 11, 2026 — Day 7 — More Room for Long Tasks
The default GPT-6.1 Sol context length in Codex and Work increased from 272,000 to 500,000 tokens. It remains configurable up to 1 million; OpenAI said subscription usage does not incur a separate long-context fee. Day 7 announcement.
October 10, 2026 — Day 6 — A Global Reset
Tibo announced a global usage-limit reset for the end of the day. Check the offer and your account's usage screen for the limits actually covered. Day 6 announcement · OpenAI reset guidance.
October 9, 2026 — Day 5 — Composer Predictions and Mobile Dots
Composer predictions entered beta for Codex desktop Pro users, suggesting the next message without consuming usage. Users could also create and text their dot in the ChatGPT mobile app. Day 5 source links.
October 8, 2026 — Day 4 — Faster Steering and an Ultrafast Mode
Instant steering was announced to help the model react sooner when a user changes direction mid-task. GPT-6.1 Sol ultrafast began rolling out in the API, Codex, and ChatGPT Work. Day 4 announcement.
October 7, 2026 — Day 3 — GPT-6 in Chat and a Banked Reset
Tibo announced GPT-6 in Chat and said a saved usage reset was being loaded into paid accounts. A banked reset must be redeemed before it expires and differs from an automatic reset; eligibility can vary by offer and account. Day 3 announcement · OpenAI reset guidance.
October 6, 2026 — Day 2 — Several Smaller Releases
Codex Auto-review became free for users signed in with a ChatGPT account, without consuming plan usage. OpenAI also simplified API usage tiers, introduced a Meetings plugin beta for eligible ChatGPT desktop users, and opened the Decisions API public beta. These releases serve different audiences; the community tracker also records a subsequent reset. Day 2 announcements and links.
October 5, 2026 — Day 1 — Faster Default Models
OpenAI said the default speeds of GPT-6 Astra and GPT-6.1 Sol increased by about 50%, from roughly 30 to 50 tokens per second across subscription products and eligible Sign in With ChatGPT partners. No user action was required. Day 1 announcement.
What to Watch Next
The updates so far mix model speed, interaction changes, product betas, and usage relief. Availability differs across plans and products, so each announcement matters more than a blanket “everyone gets everything” reading. This article covers the first fourteen days, through October 18, 2026, Pacific Time. We will add later days after they are announced and verified.
