From Anthropic to OpenAI, the industry’s elite are shifting from typing lines to managing agents—signaling a massive transformation in how software is built.
- The Zero-Code Reality: Top engineers at major AI labs like Anthropic and OpenAI report that they have completely stopped writing code by hand, relying on AI for 100% of their output.
- Explosive Productivity: With tools like Claude Code, developers are shipping over 20 pull requests a day, drastically accelerating release cycles and eliminating manual drudgery.
- A Shift in Structure: While current global adoption sits around 30%, industry leaders predict AI will handle most software engineering within a year, fundamentally reshaping open-source contributions and hiring requirements.
For decades, the image of the software engineer has been defined by the glowing screen and the rapid clatter of a keyboard—syntax, logic, and brackets painstakingly assembled by hand. However, inside the world’s most advanced AI laboratories, that era is abruptly ending. A growing cohort of elite engineers at Anthropic and OpenAI are reporting a startling shift: they simply do not write code anymore.
This is no longer about “copilots” offering suggestions or autocompleting a line. We are witnessing a transition from assistance to total dominance. At Anthropic, the creators of the Claude model, the transition is stark. Boris Cherny, Head of Claude Code, recently revealed that for over two months, 100% of his coding work has been written by AI. “I don’t even make small edits by hand,” Cherny stated, noting that he shipped 22 pull requests (PRs) in a single day and 27 the day before—a volume of output that would be physically impossible for a human typing manually.
The trend is pervasive across the organization. Anthropic estimates that company-wide, 70–90% of their code is now machine-generated. In a recursive twist of technological evolution, Claude Code now writes about 90% of its own codebase. The sentiment is echoed at rival lab OpenAI. Researcher “Roon” was blunt about the transition: “100%, I don’t write code anymore… Programming always sucked… and I’m glad it’s over.” For these engineers, the AI has liberated them from the “drudgery” of syntax, allowing them to focus entirely on architecture and intent.
While these numbers represent the bleeding edge, the gap between the elite labs and the rest of the industry is closing. Microsoft currently reports that roughly 30% of its code is AI-written, and GitHub studies show similar figures globally. However, the trajectory is exponential. Dario Amodei, CEO of Anthropic, recently warned that the industry might be only “six to twelve months away from AI handling most or all of software engineering work from start to finish.”
This shift has profound implications for the open-source ecosystem, the bedrock of modern software. Himanshu Tyagi, co-founder of the open-source AGI company Sentient, believes we are preparing for a structural change. “A large part of the code that will be shipped over the next 10 years will be written by AI,” Tyagi noted, calling Claude Code the “breakthrough product” that catalyzed this reality.
The promise for open source is significant: AI agents could lower the barrier to entry, automate tedious maintenance, and accelerate contributions. However, it also introduces complexity regarding ownership, licensing, and accountability. Furthermore, while the speed is undeniable, the technology is not infallible. Models still make “subtle conceptual errors,” suggesting that the role of the human is shifting from “writer” to “auditor” and “generalist”—someone who understands the broad strokes of a system well enough to catch high-level mistakes without getting bogged down in the implementation details.
As we move forward, the definition of a software developer is being rewritten in real-time. If the leaders of the AI revolution are to be believed, the future of coding doesn’t involve writing code at all—it involves directing the intelligence that writes it for you.