AI can scaffold your API endpoints in seconds but can't write a decent paragraph with a point of view. That's the core argument from developer Alex Wennerberg in a recent blog post that draws a sharp line between two kinds of work AI touches.
Coding: Useful for the Boring Parts
Wennerberg describes his experience with Claude as "liberatory" rather than threatening. The reason is blunt: most professional coding is what he calls "artless nonsense." Glue code, boilerplate, config files, the endless plumbing that connects systems together. AI handles this well, and handing it off frees programmers to focus on the parts that actually require thought.
But here's where it gets interesting. Rather than doubling down on AI-assisted productivity, Wennerberg went the opposite direction. He picked up assembly language and started working on reverse-engineering projects like the Pokemon Red disassembly. His argument: AI taking over the tedious parts of coding pushed him toward "recreational programming" where craftsmanship matters again.
This tracks with what a lot of developers report privately. AI coding assistants like Cursor and Claude Code are genuinely useful for rote tasks, but the satisfaction of programming comes from solving hard problems, not from generating more boilerplate faster.
Writing: AI Contributes "Almost Nothing"
The writing half of the argument is more absolute. Wennerberg claims AI contributes "almost nothing" to creative writing and is "probably even net harmful" during research and editing. His reasoning: writing exists to express something uniquely yours. Outsourcing that to a model defeats the entire purpose.
Code gets shared through open source all the time without undermining its value. But writing is personal expression. An AI-generated blog post isn't a shortcut to the same destination. It's a different destination entirely.
He draws on philosopher Maurice Merleau-Ponty to make the deeper point: human experience is embodied and particular. AI lacks a body, lacks lived experience, and therefore can't "replace" human capabilities in domains where those things are the whole point. Framing AI as a replacement for human writing is a category error.
The Practical Takeaway
This isn't an anti-AI argument. It's a sorting argument. Use AI for the mechanical work it handles well: boilerplate code, data transformation, format conversion. Stop trying to use it for work where your unique perspective is the product.
For anyone running an AI-assisted workflow, that's a useful filter. The question isn't "can AI do this task?" but "is this a task where my voice and judgment are the point?" If yes, do it yourself. If no, let the model handle it.