AI Coding Live News

Recursive self-improvement (RSI), using models to speed up the development of the next models, became one of the most-used terms in AI circles on X this past week after several big-tech CEOs talked about it in public. Google CEO Sundar Pichai framed it differently on two podcasts. On Alex Kantrowitz's Big Technology Podcast he said "We are pretty far from the end state of RSI." On Alex Heath's Sources podcast he said researchers are co-working with models "in a deep way" and "I would expect the next 12 months to see continued acceleration because of RSI." The widely shared 12-month line comes from the Sources episode, not Big Technology. OpenAI CEO Sam Altman told Stanford GSB that besides AI recursive self-improvement there is another loop, robots that build robots and data centers that build data centers, that gets "much less talk". Anthropic CEO Dario Amodei, replying to a critic on X, wrote about the point where AI-driven acceleration gets fast enough that safety struggles to keep up: "We are nearing that point." This is an X discussion of public remarks by CEOs, not an announcement of any new research program.

Following Dario Amodei safety essay, Sam Altman, Elon Musk, and Demis Hassabis engaged in public dialogues exploring a Frontier Pacing protocol. The proposed voluntary mechanism would trigger a 90-day external security audit buffer whenever next-gen models breach critical capability thresholds.

OpenAI Codex & ChatGPT lead Tibo says capacity scaled better than expected, so GPT-6 Astra is now rolled out to all Plus and Business users (after Pro / Enterprise / Business Premium and the API). Sam Altman followed up confirming it is out to all Plus and Business users. A key widening of Astra access beyond premium tiers for ChatGPT and Codex subscribers.

OpenAI is preparing to release Astra, saying the model’s cybersecurity capability has reached the Critical threshold under its Preparedness Framework. Sam Altman confirmed Astra has finished training and is a significant step in both capability and alignment, while later models are being paced so safety work can keep up. OpenAI is previewing its evaluations, safeguard advances, and remaining open questions.

Sam Altman announced that OpenAI paused certain frontier reinforcement learning training runs where capability growth outpaced alignment and safety tooling, while confirming near-term model releases remain on schedule.

OpenAI rolled out GPT-5.6 Sol API preview featuring diagram-to-code synthesis and automatic o3-mini hybrid routing.

Sam Altman announced that OpenAI has paused some frontier reinforcement learning training so alignment, security, and monitoring can catch up with a new level of model capability. He emphasized that safety confidence will increasingly set the pace of AI progress.