Every reader as a market of one.

Headless Architecture: A Love Letter

In data-driven software systems, technology developers often face a fundamental architectural choice: tightly couple content and presentation or keep them separate.

Some systems are “monolithic.” In this architecture, the data on the server and what we see in apps and on websites are tightly coupled. Other systems are “headless.” In these systems, the data on the server and what we see in apps and on websites are separate or “decoupled.”

Each architecture has its advantages, but headless architectures can be extraordinarily flexible. Underlying data and the way it’s presented can evolve independently. Neither is trapped or limited by the other. Headless allows data “primitives” to be presented and consumed in any number of different ways. One server-side data layer can easily become a series of reports, feed an application, add content to a website, and facilitate intelligence audits.

By decoupling the underlying atomic data from any single presentation layer, we liberate the value of the former from the restrictions of the latter. Abstraction becomes a virtue. And no single container artificially limits the value of the underlying asset.

The specific core components of a headless system follow:

• data lives in a central repository (like a server-side database)
• APIs serve as transmission lines between the server and the browser, app, website, or other user-consumed asset on the “front end”
• a frontend browser, app, or website renders the API into something people want to read or use

What happens when we apply these components to traditional writing?

Headless Writing or: How I Learned to Stop Worrying and Love Literate AI

Since literacy began, writers and texts have been almost exclusively monolithic. Writer, voice, text, and utility are part of a single mass: a law code written by a sovereign frames society, a scholarly monograph written by a subject-matter expert makes a topic legible to readers, a creative work with a particular style entertains fans.

Monolithic writing has served human beings well for thousands of years. And it will always have a role. However, the relationship between writers and readers has substantially changed in the past 35 years. Writers tend to still want their insightful, beautiful, one-of-a-kind monoliths: their perfect essays, short stories, and think pieces. But readers now want 15-second media: short videos and posts, bottom-line-it-for-me, anti-TLDR summaries.

This disconnect between creator and consumer is compounded by the fact that there are far more things to read now than there used to be. There are more people creating more content, and most of the traditional capital barriers to audience access and content distribution have collapsed. Compounding this, we’re now in a situation where AI reads and writes better than human beings. We have a new, ultra-competitive intelligence capable of producing and consuming even more content. Taken together, this means the old economics that used to bind readers and writers together no longer makes sense. Value can now accrue to different kinds of systems native to this new world.

Using a headless-inspired system for writing resolves this complicated state of affairs in some interesting ways.

Here’s the model I have in mind:

• The monolithic text survives as the central “database.” Human beings are very good at creating these kinds of assets. We are our stories. We are our law codes. We are our timeless essays. We are our poems and plays. Human beings own this layer and keep writing just as we’ve always done.
• Document atomization becomes the structured content layer: literally, as a JSON representation of “rhetorical atoms” that software can surface as an API and use to recompose new texts and, figuratively, as the transmission layer between writer/monolith and what the reader actually consumes. AI facilitates this process.
• We then use these atoms to create purpose-built geometries for particular reader personas, which we wrap in an appropriate genre these readers want to read. AI enables this.

In this model, we can isolate the key rhetorical primitives from a single textual monolith and transform these into emails, memos, creative works, and other forms offered to individual readers we treat as markets of one.

Four cornerstones

1. First, while no atomization is completely lossless, monolithic documents have underlying semantic/rhetorical/style structures that can be separated from any particular formal expression.
2. Second, AI makes it economically feasible to extract that structure, organize it, and reassemble it repeatedly in meaningful ways.
3. Third, once the first two things are true, the traditional document stops being the only possible unit of authorship and readers can consume the source optimized specifically for them.
4. Fourth, these terms are distinct: “Document atomization” is the process. “Rhetorical atoms” are the structured content (often expressed as JSON). And “headless writing” is the top-level architecture.

The qualitative goals of headless writing and document atomization are also simple:

We’re not simply chopping text into pieces. In classic headless fashion, we identify reusable structure beneath a particular monolith and separate semantic content from presentation geometry. One rhetorical substrate can support many audiences and many genres.

This isn’t just omnichannel publishing where one monolith is distributed to many channels. This is semantic omnichannel writing where one monolith becomes many different forms tailor-made for many different readers.

In headless writing, the monolith doesn't disappear. It becomes the source of truth. Document atomization separates meaning from presentation so the same underlying ideas can be rebuilt for a particular reader, channel, and moment.

In 1967, Roland Barthes attacked a literary tradition that privileged the Author, arguing that, “to restore to writing its future … the birth of the reader must be ransomed by the death of the Author.”

Now, in 2026, looking out to a future with literate AI, we might extend Barthes’ vision of the future of writing to include “the atomization of the Text.”