The person who built the tool eating software jobs says the feast is just starting.
The Summary
- Boris Cherny, lead engineer behind Claude Code at Anthropic, predicts major software engineering job losses from AI automation while also forecasting new job creation in areas humans still dominate.
- The person building the automation is unusually candid: this isn't hype, it's happening, and the timeline is compressed.
- The implication: if you're still writing boilerplate in 2026, you're already competing with a tireless $20/month colleague.
The Signal
Boris Cherny doesn't work in the abstract future. He builds the tools remaking software work right now. As the lead engineer on Claude Code, Anthropic's AI coding assistant, he has front-row seats to what happens when agents get good enough to ship features, not just suggest autocomplete. And his read is blunt: major job displacement is coming for software engineers.
This matters because Cherny isn't a futurist with a book deal. He's writing the code that automates code-writing. When he says the shift is real, it carries weight. The typical discourse around AI and jobs splits into two camps: breathless techno-optimists who think everyone will be a CEO of one, and skeptics who point out that autocomplete isn't intelligence. Cherny occupies the pragmatic middle: yes, automation will crater demand for certain engineering roles. Yes, new jobs will emerge. No, the transition will not be smooth.
"The people building the automation tools are the ones with the clearest view of what's actually getting automated."
The distinction Cherny makes is critical. AI won't replace all software engineers. It will replace the ones doing work that's already been solved a thousand times before. CRUD apps. API wrappers. Boilerplate factory after boilerplate factory. The economic logic is simple:
- A junior engineer costs $80K-120K annually in most markets
- Claude Pro costs $20/month and works 24/7
- The gap between "can't ship anything" and "ships simple features with oversight" closed faster than anyone budgeted for
Meanwhile, the work that remains human-shaped gets more interesting and more demanding. You need engineers who can architect systems, understand business context deeply enough to ask the right questions, and debug the weird edge cases agents still flub. The floor rises. The ceiling stays high. The middle hollows out.
The Implication
If you're building software in 2026, the question isn't whether AI will change your job. It's whether you're developing skills agents can't replicate yet. System design, taste, the ability to translate messy human problems into technical architecture: these still require human judgment. Grinding out forms and API glue code does not.
The jobs emerging on the other side of this transition won't look like traditional engineering roles. They'll blend technical fluency with domain expertise, product sense, and the ability to direct autonomous systems. Start building that skillset now, or start competing with tools that cost less than your coffee budget.