What’s one thing that’s missing in codex that you wish we had?

ACR Poker Update Regarding TrenirusGTO Following our statement earlier today regarding the video showing TrenirusGTO being used during live play on ACR Poker, our Security and Game Integrity team has identified the account involved. The ACR Poker username shown in the video was UGLYRETD. Our investigation determined that the account was created on 8/2/2026, deposited $14.13, and played on ACR Poker for a single day. During that time, the account played 65 hands at $0.05/$0.10 stakes, 27 hands at $0.02/$0.05 stakes and 19 hands at $0.01/$0.02 and recorded a net loss of $-13.12. The account has been permanently banned from ACR Poker. While the account was a net loser overall, it did win individual hands against other players during the session. A total of 7 players who lost money directly to this account will be refunded for the affected hands and notified within the next 24 hours. Our Security and Game Integrity team worked quickly to identify the account, review the relevant play, and establish the scope of the activity after the video was brought to our attention. We also want to thank the poker community for bringing this matter to our attention and for the information and support provided throughout the investigation. Protecting the integrity of our games is a shared responsibility, and reports from players and the wider poker community are an important part of that process. As new forms of prohibited assistance continue to emerge, we will continue evolving our detection, investigation, and enforcement processes. We will not publicly detail those methods, but our position remains unequivocal: Real-Time Assistance is prohibited on ACR Poker. Accounts found using it will be permanently banned, and affected players will be made whole. We appreciate the poker community holding us accountable, and we will continue working to make ACR Poker a safer and fairer place to play. - ACR Poker

We are working towards getting Astra in everyone's hands as quickly as we can; I know it is frustrating and I appreciate the patience. It should be quick.

🚨 OpenAI have launched a waitlist for 5.6 Sol Ultrafast, a version of Sol that streams at up to 750tps - 14x faster than standard

anthropic already has the model after Mythos 5. it is internal, and you may never get to use it. buried in Anthropic’s 186-page risk report is “Model 2,” an internal only system Anthropic says is a noticeable improvement on Mythos 5 in many ways. on CoBench v2, Model 2 is dropped into a past version of Anthropic with old code, logs, internal messages and documents, then asked to diagnose 449 real training and infrastructure problems. it scored 62.8%. Mythos 5 scored 50.3%. Anthropic estimates roughly 85% would mean a model could fully substitute for its technical staff and could mean RSI. Model 2 and Mythos 5 already run as persistent research agents. Claude now writes a large majority of code merged into production. the lab says AI has made its R&D significantly faster, though still by less than 2×. anthropic.com/aug-2026-risk-… Anthropic currently has no plans to release Model 2. this is how the public frontier falls behind the private one. Anthropic’s researchers are already using one or more models ahead, using that model(s) to help build the next model.