March 8, 2026
When code joins a book club
We should revisit literate programming in the agent era
“Write novels for robots?” Devs clash as AI revives story‑style code
TLDR: A developer argues AI assistants now remove the grunt work from “code as narrative,” making literate programming practical again. Commenters split between “finally, clarity wins,” “code is the only truth,” and “funny how we’ll write docs for robots,” highlighting a culture clash with real stakes for how teams build software.
A developer says the old idea of literate programming—mixing code with plain‑English explanations—finally works now that AI assistants can do the boring parts. Think Jupyter notebooks or Org Mode files that read like a how‑to guide, where an agent handles re‑exporting code and keeping prose and code in sync. The pitch: let AI tidy the paperwork so humans focus on ideas.
Cue the comments section meltdown. One camp cheered, arguing that if bots like this thrive on language, then clear writing pays big dividends. Another went full skeptical: user sublinear snapped that engineers have always spent more time reading than writing, and warned that too much prose “inevitably misrepresents the actual code.” Translation: storytime can lie. A practical middle ground emerged from perrygeo—skip the novel, do solid names, short comments, and a great README. Meanwhile, rustybolt dropped the spicy truth-bomb: folks wouldn’t write for humans, but suddenly they’ll write for the bot. The meme of the day: “Do it for the clanker.”
Others proposed a truce: build strong configuration layers and let agents fill in the blanks—less magic, more reviewable output. And the purists? They rolled their eyes: “Isn’t the code the documentation?” Whether this is a comeback or a book club for code, the debate is on. Read the literate programming origin story, and bring popcorn.
Key Points
- •Literate programming interleaves code and prose to make systems understandable as a narrative.
- •Jupyter notebooks exemplify literate programming by combining explanations with calculations and outputs.
- •Emacs Org Mode and org-babel enable polyglot literate programming with executable code blocks and captured results.
- •Maintaining parallel code and prose and handling tangling have limited broader adoption, especially in larger projects.
- •Coding agents (e.g., Claude, Kimi) can generate Org runbooks, update code and prose, handle tangling, and reduce maintenance overhead.