The creator of Claude Code predicts the end of the software engineer. What's behind Anthropic's agent-based coding tool — and what companies should do now.
The End of the Software Engineer?
Boris Cherny, chief architect of Anthropic's Claude Code, put it bluntly: "The title Software Engineer will disappear." Instead, everyone will become a "Builder" — supported by AI agents that write code autonomously. A bold claim. But one you have to take seriously when it comes from the creator of one of the most powerful AI coding tools available today.
Cherny himself says he hasn't written a single line of code by hand since November 2025. Everything runs through Claude Code. And he's convinced: In one to two years, the traditional fundamentals of software development will no longer matter.
Provocative? Certainly. But also a signal you shouldn't ignore.
Klingt interessant?
What Is Claude Code?
Claude Code is what's known as an agent-based AI coding tool from Anthropic. Unlike classic code assistants like GitHub Copilot, Claude Code takes a decisive step further:
- It executes complete software development tasks autonomously
- It plans, implements, and iterates — with minimal human intervention
- It doesn't just work at the file level but understands entire project structures
The concept behind it is called Agentic Coding: The AI is no longer a tool waiting for instructions. It's an independent agent that accepts tasks, makes decisions, and delivers results.
"I haven't edited a single line of code by hand since November." — Boris Cherny, Chief Architect of Claude Code
What Anthropic Itself Says: AI Is Taking Over Software Development
Anthropic likes to position itself as the "responsible" AI company. Which makes it all the more remarkable how clearly Cherny describes the future of software development:
- Everyone becomes a Product Manager. The ability to define problems and formulate requirements becomes more important than technical implementation
- The title "Software Engineer" disappears. It's replaced by "Builder" — people who collaborate with AI agents instead of writing code themselves
- The disruption comes fast. Cherny acknowledges that the transition will be "painful for many"
At the same time, he qualifies: Claude Code still has limitations. You need to verify the results, ensure the code is correct and secure. Fully hands-off isn't possible yet.
The honest assessment: We're in a transition phase. The AI is good enough to handle the bulk of implementation. But not good enough to work without human oversight.
What This Means for Developers and Teams
The Uncomfortable Truth
If a tool like Claude Code can write 100% of the code in the hands of an experienced engineer, the question arises: What happens to the teams that currently deploy 10 developers for a project?
The answer is uncomfortable but clear:
- Fewer developers will produce more output. A team of 3 with AI agents can deliver what 10 are needed for today
- Junior positions will change radically. The entry via "simple" implementation tasks disappears when AI handles them faster and cheaper
- Routine coding becomes a commodity. CRUD operations, boilerplate, standard integrations — all automatable
The Job Market Is Already Reacting
The tech industry is feeling the pressure right now. Companies are hiring fewer traditional developers and investing instead in AI-powered workflows. This isn't a future scenario — it's happening now.
Opportunities: What Stays Human?
Despite all the disruption, there are areas where human expertise doesn't just remain relevant — it becomes more important:
- Architecture decisions — What systems do we need? How do we scale? What trade-offs do we accept? This requires experience and judgment
- Domain knowledge — AI can write code, but it doesn't understand why a Swiss insurance process works differently than an American one
- Quality assurance — Code reviews, security audits, and architecture reviews become more important, not less
- Stakeholder communication — Understanding requirements, managing expectations, translating technical concepts — that remains human work
- Ethics and responsibility — Who is responsible when AI-generated code has a bug? A human must answer that question
The shift is clear: From "writing code" to "orchestrating code." Those who understand this change have an advantage. Those who ignore it have a problem.
Recommendations for Companies
1. Evaluate Now, Don't Wait
Technology is evolving faster than most companies can react. Anyone who starts testing AI agents only in 12 months has already fallen behind.
2. Redesign Workflows, Don't Just Add Tools
Slapping an AI coding tool on top of existing processes yields little benefit. The real leverage lies in redesigning the entire development process:
- Professionalize requirements engineering — AI needs clear requirements
- Strengthen review processes — More code means more review is needed
- Expand test automation — Without tests, AI-generated code is a risk
3. Redefine Roles
Instead of "frontend developer" and "backend developer," teams will need:
- AI Architects — who define how AI agents are optimally deployed
- Quality Engineers — who ensure the output is correct
- Product Engineers — who connect business logic with technical implementation
4. Invest in People
Paradoxically, the human component doesn't become less important through AI — it becomes more valuable. Invest in:
- Architecture know-how
- Domain expertise
- Communication skills
- Strategic thinking
Conclusion
Boris Cherny's prediction isn't clickbait. It's a well-founded assessment from someone who built one of the most powerful AI coding tools in the world. Whether the title "Software Engineer" actually disappears is secondary. What matters is that the role is fundamentally changing.
Companies that act now will have a decisive edge in 12 months. Not because they'll need fewer people — but because their people will work alongside AI agents and achieve multiples of what they could before.
The question isn't if, but when. And "when" is now.
