Defining the new debate: Is coding already “solved”?
The debate over AI replacing software engineers centers on whether advanced coding agents can handle most programming tasks so efficiently that the traditional engineering role becomes less central and more people focus on defining problems, products, and user needs while AI writes much of the code. Boris Cherny, creator and head of Claude Code at Anthropic, sits squarely in this camp. He has not written a line of code himself in more than six months and says that, for the type of work he does, coding is effectively “solved.” He predicts that the job title “software engineer” could start to disappear, replaced by broader labels such as “builder” as designers, product managers, and non‑technical colleagues use tools like Claude AI to ship production code on their own. This marks a sharp break from past assumptions that professional developers are always the main gatekeepers of software creation.
Claude AI capabilities and the shrinking moat around coding
Claude AI, and especially Claude Code, shows how AI impact on developer careers is accelerating. Cherny describes the tool as an agentic coding system that can take high‑level instructions and turn them into working software, reducing the need for humans to type every line. In a recent session with the latest batch of Y Combinator founders, he asked how many let Claude Code write “100% of their code” and saw about half the hands go up. Only one person in a room of a couple hundred said they wrote all code without the model. These numbers suggest that for a growing share of greenfield projects, AI replacing software engineers’ routine tasks is no longer speculative. Instead, the scarce skills shift from syntax mastery to reviewing, testing, and defining what should be built in the first place.
From software engineer to builder: which roles are at risk?
If coding itself becomes a commodity, the software engineering jobs future will depend on which parts of the role AI cannot absorb. Tasks that follow clear patterns—CRUD apps, boilerplate integrations, straightforward refactors—are the most vulnerable, as Claude Code and similar tools can generate and maintain that code quickly. Entry‑level roles that once focused on these repetitive tasks may be hardest hit. By contrast, roles that blend technical understanding with product sense and leadership gain importance. Cherny expects the title “software engineer” to fade, but not the work of creating software. He anticipates many more people writing code or directing agents to write it, including designers and product managers who already own key decisions about user experience. Quality assurance, security review, complex systems design, and ethical oversight remain difficult to automate and may become the new core of high‑value technical careers.
“Golden age” for CS grads and founders despite disruption
For new computer science graduates, Cherny’s message is strikingly optimistic. He tells 22‑year‑old CS grads that if they are at all entrepreneurial, this is the time to found a startup. Thanks to tools like Claude Code, small teams—or even solo founders—can build and scale products that once required large engineering departments. Cherny calls this “the golden age” and predicts that, as AI lowers the barrier to building, there will be “100 times more” people writing code or using agents to write code than today. This fits with comments from Sam Altman, who said he now wants to fund founders who deeply understand users even if they “can’t code at all.” For graduates, the opportunity lies in pairing AI coding agents with domain insight, rather than competing head‑on with machines on raw implementation speed.
Planning a career in an AI‑first software world
For anyone worried about AI replacing software engineers, the lesson from Anthropic’s Claude story is not to abandon software, but to rethink where to specialize. Careers anchored only in routine implementation are more fragile; careers built around problem framing, architecture, system reliability, and user understanding are more resilient. Learning how to collaborate with tools like Claude AI, instead of ignoring them, is becoming essential. Cherny’s own behavior—automating his job and letting agents write code—shows how far this shift has already gone at the frontier. The safest long‑term path is to treat coding as one ingredient in a broader craft: defining problems, validating solutions, and orchestrating AI systems and human teams. In that world, the job title may change, but skilled builders, reviewers, and product thinkers remain in high demand.