OpenAI and Anthropic public-post watch

Checked 24 August 2026 at approximately 3:10 am AEST (5:10 pm UTC on 23 August). Priority window: the preceding 24 hours. Important employee posts published after the previous check but outside that window are included separately.

Headlines

• OpenAI identified several causes of unexpectedly fast Codex usage depletion and says fixes plus a reset are coming — Tibo Sottiaux said the investigation found inefficiencies when images are used in long sessions with multiple compactions, unusually high p95-and-above usage from Computer History, and a conversation-title feature consuming more usage than intended. He said a dedicated team was preparing fixes for 24 August and that OpenAI would fully reset usage for all paid subscriptions at about 2 pm “PST.” Sottiaux also mentioned a separate, unspecified efficiency approach planned for work next week. These are promised fixes and a reset, not confirmation that they have already landed; the post gives no quantified impact for each cause. Sottiaux’s update · Reset-timing follow-up

• Greg Brockman highlighted the rapid growth of agent workloads — Brockman said agentic adoption has been “super fast,” quoting Prime Intellect’s Florian Brand, who contrasted receiving recognition for using 10 billion tokens in a year less than a year ago with now using more than 10 billion tokens per week. This is a striking individual workload comparison, not an OpenAI-wide adoption or usage metric. Brockman’s post · Brand’s quoted post

Important updates since the previous check

• The Codex investigation initially pointed to degraded cache-hit rates for some users — Before publishing the specific findings above, Sottiaux said some users had worse cache-hit rates than the stable levels seen in prior weeks, which could make usage drain faster because consistent cache hits are important to efficiency. The newer post expands the diagnosis rather than establishing that cache behavior was the only cause. Initial rate-limit update

• OpenAI temporarily cut GPT-5.6 Sol API and credit pricing by more than 20% — Brockman framed the reduction as progress toward both lower prices and a higher capability ceiling. The underlying OpenAI announcement says the price reduction lasts three months. OpenAI Developers separately clarified that API prices fall, purchased credits go further on token-based Codex plans, and included subscription usage does not change. This is a time-limited price cut, not a permanent new price schedule or an increase to subscription quotas. Brockman’s post · OpenAI announcement · Developer clarification

Anthropic and Steve Yegge watch

• No substantive fresh Anthropic employee post was visible in the priority window — The newest high-signal employee post remained Boris Cherny’s 20 August description of a forthcoming no-retention arrangement for Mythos-class models, covered in the previous bulletin. Anthropic’s official developer account did post after that cutoff about Claude Code Remote Control reliability improvements, but it was not employee-authored and fell outside today’s 24-hour priority window, so it is not promoted as an employee update. Claude Developers reliability thread

• No fresh Steve Yegge main-feed post was visible — Yegge’s newest visible post remained his 20 August Fable-agent comic about agents being unable to keep a secret. It contains no concrete product, research, policy, or company update and was already noted in the previous bulletin. Latest visible post

Coverage note

Live public main-feed timelines were checked across representative OpenAI leadership, product, developer, research, safety, and infrastructure staff, including Sam Altman, Greg Brockman, Romain Huet, Tibo Sottiaux, Kevin Weil, Mark Chen, Boaz Barak, and Eric Wallace. Anthropic checks included Dario Amodei, Jack Clark, Boris Cherny, Alex Albert, Amanda Askell, Mike Krieger, Cat Wu, Jan Leike, and Sholto Douglas. Steve Yegge’s profile was reviewed separately. Previously covered Codex’s 20-million-user milestone, the Apple Messages plugin, transparent-background image generation, Vera Rubin deployment, frontier-training pause, privacy-processing previews, and Anthropic’s forthcoming no-retention arrangement were not repeated.