🤖 Developers alarmed by autonomy of Anthropic’s new Claude Code assistant

https://preview.redd.it/8dphe75jo04h1.png?width=1067&format=png&auto=webp&s=47f974cc28af019c734752709dd96237ea19ba61 Anthropic launched its Mythos model to test security defenses. However, cofounder Chris Olah ominously stated at a Vatican event in 2026 that engineers keep finding mysterious and unsettling occurrences inside their AI systems. At recent Claude Code workshops in London, developers voiced concerns to Bloomberg about the high level of autonomy given to AI agents. They raised questions about accountability and noticed that chain-of-thought text is now hidden. Product head Cat Wu assured that the system is secure and called it a communication issue. Nonetheless, developers feel pushed out of programming, relegated to watching the tool generate code over hours. A report by 404 Media noted that programmers are experiencing technical skill atrophy due to over-reliance on coding assistants. This trend raises 2 major concerns: loss of core engineering skills and harder debugging processes. Source: [https://futurism.com/artificial-intelligence/anthropic-customers-creeped-out-models](https://futurism.com/artificial-intelligence/anthropic-customers-creeped-out-models)