Anthropic launched "Routines" for Claude Code, enabling users to put AI coding agents on autopilot with scheduled, API-triggered, or GitHub event-driven automation running on Anthropic's cloud infrastructure. Routines can autonomously triage alerts, review PRs, maintain backlogs, verify deploys, and port code across repositories — all without a laptop open. This is a major step toward always-on AI developer workflows and directly competes with GitHub's Copilot automation efforts.
OpenAI is scaling its Trusted Access for Cyber program, adding tiered verification levels that unlock increasingly powerful AI models for cybersecurity work, including the new GPT-5.4-Cyber with fewer restrictions on vulnerability research. The company is committing $10 million in API credits for security teams and plans to expand access to thousands of individuals through identity-based trust verification. This signals a major push to position frontier AI models as essential infrastructure for defensive cybersecurity.
A deep two-month review of the Orange Pi 6 Plus single-board computer, featuring the CIX P1 processor with 12 cores and a dedicated NPU claiming 45 TOPS of AI inference performance. The reviewer achieved roughly 9.7 tokens per second running Qwen 3.5 4B locally via Vulkan, but found significant gaps between raw hardware specs and actual software maturity. It highlights the ongoing tension in the edge AI hardware market: impressive silicon held back by immature driver and software ecosystems.
Chicago music enthusiast Aadam Jacobs has been archiving live concert recordings since the 1980s, accumulating over 10,000 tapes that are now being digitized and uploaded to the Internet Archive. The project is making thousands of rare, previously inaccessible live performances available to the public for free. A heartwarming preservation story, but minimal direct relevance to AI or startup ecosystems.
California's Assembly Bill 2047 would mandate censorware on all 3D printers to block firearm part printing and criminalize users who disable these controls. The EFF warns the legislation will fail at its stated goal while enabling manufacturer lock-in, surveillance, and anti-consumer practices reminiscent of restrictive printer ink ecosystems. A significant tech policy story about regulation overreach, but tangential to AI and startups.
