The MIT Technology Review didn't soften the framing: Anthropic's Code with Claude event showed off coding's future - and the "whether you like it or not" qualifier was doing real work.
The May 21 showcase put Claude Code front and center. Anthropic's AI coding agent takes a plain-English description of what software should do and writes working code, runs tests, catches errors, and fixes them - without a human approving each step. That's a different category from GitHub Copilot-style autocomplete. It's closer to handing a task to a junior developer and getting back a pull request.
That's the part some developers aren't comfortable with. Not because the demos failed - because they succeeded.
The friction at events like this isn't really technical. It's professional. When a coding tool moves from "helps me write boilerplate faster" to "completed this feature ticket while I was in a meeting," the job description starts to shift. Developers who spent years getting good at exactly the things Claude Code now handles competently are being asked to redefine what their value is.
For non-developers who use AI tools to build things - marketers who want a custom dashboard, freelancers automating their invoicing, small business owners who need a simple internal app - this trajectory is mostly good news. Claude Code keeps getting more capable at the kinds of tasks that used to require hiring someone or knowing Python.
For working programmers, the calculation is harder. The same tools that make individual developers more productive also make individual developers more substitutable. Both things are true at the same time, and showcases like this one make that tension harder to look away from.