Les Orchard's March 2026 blog post hit the front page of Hacker News this week, pulling in hundreds of comments from developers who said they recognized something they'd never had language for: a professional split that predates AI tools but only became consequential once those tools arrived.
Orchard, who has been programming since a Commodore 64 in the 1980s, argues two fundamentally different developer types have always coexisted — those who code for the craft itself, and those who code to make things happen. They used the same tools and workflows, so the difference was invisible. AI coding assistants like Anthropic's Claude Code created a fork in the road that made hidden motivations suddenly visible and suddenly divisive.
To build his case, Orchard draws on a range of named voices. Nolan Lawson's essay "We Mourn Our Craft" captures what craft-oriented developers are losing — the sculptor's satisfaction of hand-molding code, the pride of authorship, the 2 AM debugging triumph. Developer James Randall describes a "compressed" sense of discovery that AI tools bring. Orchard identifies with the make-it-go camp and reports that his feared losses — losing the ability to judge AI output, losing the puzzle-solving satisfaction — largely didn't materialize. The puzzle, he writes, simply shifted up a level to architecture and direction, consistent with every prior transition in his career. Simon Willison endorsed this framing in the HN thread, linking it to Kellan Elliott-McCrea's longstanding argument that "code has always been the easy part" and that real value in software resides in the human-technology system.
Orchard's own grief isn't about writing code. It's about the world around it: the <a href="/news/2026-03-14-optimizing-web-content-for-ai-agents-via-http-content-negotiation">open web ecosystem being consumed by AI training on the commons</a>, and a structural shift in developer demand away from traditional web work toward AI engineering. The HN thread surfaced different objections. Several top-voted comments pushed back on the craft-vs.-purpose framing — one argued the real divide is between developers who believe software is fundamentally understandable and those comfortable treating it as a black box; another applied a labor lens, noting that productivity gains historically translate to more work rather than better conditions, and an AI-powered productivity tool doesn't change that dynamic.
Those objections point at something Orchard's framing, for all its clarity, doesn't fully address. The developer identity question is entangled with who profits from <a href="/news/2026-03-14-nyt-ai-coding-assistants-end-of-programming-jobs">AI-accelerated output</a> and who owns the commons being consumed to train the models. The craft-vs.-purpose split describes the loss. It doesn't say anything about who's collecting the upside.