Writing and AI

These pages ask what happens when writing stops being imagined as a solitary, linear act. The Work of Writing in the Age of AI sets the historical frame in three stages. The first uses sharp objects on surfaces — oracle bones, cuneiform, papyrus — and shares its technical base with painting. The second arrives in the 1870s with the typewriter and continues into the keyboard, and brings with it “the dramatic reduction in the effort required for training” along with the model of the immaculate, monospaced page. The third, the essay argues, is now: AI-collaborated writing, in which “human ideation” combines with “the wordsmith power of language models.” The point is not that the third stage is obviously better, but that it changes which steps of composition belong to the writer and which can be shared.

The essay’s polemical move is its defense of mess. “Our thoughts are not cleanly delineated. They are a tangled, messy network of interconnected ideas. Writing is about shaping this tangled mess, or to put it in more accurate terms, about helping this mess to take shape by itself. The messy look of a manuscript therefore is a faithful picture of our mental work.” From there it argues against the minimalism cult of distraction-free apps — the empty page as moral discipline — and for tools that admit clutter as part of thinking. The pictures of Susan Sontag’s room (and the missing one of Duras’s) are not anecdotes; they are evidence of what a writer’s environment actually looks like. Once the third stage arrives, the question becomes whether AI-assisted writing tools will preserve or efface that environment.

Tabwriting and the Future of Writing Technology sets up the longer media-historical context. Writing on a slate with a stick is not a primitive antecedent to pen and paper; it is a recurring form, from oracle bones and cuneiform to children’s chalkboards and the iPad. The essay reads tabwriting as a return to a much older affordance — direct mark-making on a hard surface — and argues that contemporary touchscreens are “not envisioned for writing, but rather, for the effective recognition of brief gestures.” Until points and swipes are converted into strokes at the system level, handwriting on tablets will remain a niche solution. The lesson the essay draws on is Marcel Mauss’s notion of body technique: handwriting is not a free-standing skill but a tradition transmitted through “material aspects (pen-hold, copy books) as well as mental aspect (teaching school and human guidance).”

Mindcraft is the site’s own working answer — a next-generation writing platform “where human collaborate with AI,” built on top of a Notion-style WYSIWYG editor with extended functionality for sidebars and specialized content presentation. The project is at an early stage; what matters here is the design intent. If thinking under AI is more dialogic, provisional, and environment-shaped, the writing tool has to look less like a typewriter (one stream of immaculate text) and more like a workshop (multiple panes, drafts in conversation, AI as a participant rather than an autocomplete).

What emerges across these pieces is not just a story about efficiency. They argue that AI changes both the practice and the software form of writing — making composition more conversational, more iterative, and more obviously shaped by the environments in which thought gets organized. The connection to Harness Engineering is direct: the question of how to write with AI is the writing-side version of the question of how to build with it.