In August 2025, a meetup called "Claude Code Anonymous" took place in London. Organizer Peter Steinberger gathered a group of self-described "addicts"—tech professionals united by their fascination with Anthropic’s groundbreaking coding tool: Claude Code Opus 4.5. This AI agent handles far more complex programming tasks, boasts a dramatically expanded memory, runs for hours on end, and can even manage a team of multiple sub-agents. Anthropic claimed that in its notoriously tough internal engineering hiring test, Opus 4.5 scored "higher than any human candidate," a comparison that raised serious questions about the future of engineering careers.
By November of the same year, Steinberger launched a tool called OpenClaw, which lets users easily create their own AI agents, leveraging the latest capabilities of Claude Code or other coding tools. OpenClaw marked the shift of AI agents from professional labs to the desks of individual developers. When these agents can autonomously execute multi-step tasks, the unit of value moves from a single inference to a full deliverable, shattering the traditional API billing model. Companies were forced to reevaluate development processes, and developers faced the reality of their roles being compressed or even replaced—this shockwave quickly spread beyond code generation, leaving the entire tech industry in disarray.
