Pseudonymous Bitcoin veteran CobraBitcoin warned that increasingly autonomous artificial intelligence could uncover critical vulnerabilities in Bitcoin Core, Lightning implementations, wallets and libraries as AI agents become broadly accessible. OpenAI released GPT-6 Astra on September 3, 2026, after training it on more than 100,000 GPUs, and reported scores of 99.9% on ARC-AGI-3, about 97.6% or 98% on FrontierMath Tier 4, 100% on ExploitBench, 42.4% on ExploitGym, 57.9% on Terminal-Bench 4.0 and 59.3% on Agents’ Last Exam. Astra found and used two zero-day exploits in an internal test and became the first OpenAI model to exceed the company’s "Critical" cybersecurity-risk threshold. Access began with selected cybersecurity organizations through the Daybreak program before expanding to ChatGPT Plus, Pro, Business, Enterprise and API users, although a product-page update described the broader rollout as occurring over the following days. The model’s recurrent-depth inference and autonomous coding capabilities are expected to increase demand for accelerators, high-bandwidth memory and high-speed networks; OpenAI’s next model is planned to use 400,000 graphics processing units.