Sam Altman apologizes for GPT-6 Astra lockout as Jensen Huang declares AGI has arrived
An unstable rollout locked out paying ChatGPT Plus and Pro subscribers, overshadowing high praise from NVIDIA’s CEO.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- OpenAI CEO Sam Altman issued a public apology after the GPT-6 Astra rollout locked out thousands of paying users.
- NVIDIA CEO Jensen Huang praised Astra's agentic computer-use capabilities, declaring that artificial general intelligence has arrived.
- The server lockout sparked criticism over OpenAI's engineering reliability during high-concurrency flagship releases.
OpenAI’s launch of its next-generation flagship model, GPT-6 Astra, has encountered immediate deployment hurdles, prompting a public apology from CEO Sam Altman. Within hours of the model's rollout to professional and developer tiers, a severe server concurrency failure locked out thousands of paying ChatGPT Plus, Pro, and Enterprise subscribers, preventing them from accessing both the new model and existing services for several hours.
The technical disruption overshadowed a ringing endorsement from NVIDIA Chief Executive Officer Jensen Huang. Speaking at a technology summit, Huang congratulated OpenAI and declared that Astra's integration of native desktop control and complex software reasoning signals the official arrival of artificial general intelligence (AGI). Huang praised the model's ability to operate real-world software packages like KiCad and Blender as a watershed moment for autonomous digital agent deployment.
Altman characterized the deployment failure as a 'messy' and frustrating experience for OpenAI’s most loyal customer base, attributing the lockout to a database synchronization bottleneck in the model's new persistent note-taking architecture. OpenAI engineers implemented emergency rate-limiting and rolled back the fast-mode routing layer to restore basic service. The incident has intensified criticisms from developer forums, with many pointing out that OpenAI's infrastructure continues to buckle under the high concurrency demands of major product launches.
This rollout turbulence illustrates that scaling the physical infrastructure for autonomous agentic systems is as much a bottleneck as training the models themselves. While NVIDIA’s CEO is declaring the arrival of AGI based on Astra’s logical reasoning, the server-side failure shows that deploying continuous, window-monitoring AI agents places an extraordinary operational load on web servers, forcing developers to balance theoretical model capabilities against real-world reliability constraints.
Will OpenAI disclose a detailed post-mortem explaining the server synchronization failures that caused the subscriber lockout?
Still open. When the paper finds out, it will say so here and on the open questions page — including if it got this wrong.