The Narrative is the Business
Mar 31, 2026
All work is converging on writing. What does it mean for a business to write to the market, to its team, to its agents? And what does it hear back?
All work is converging on writing. Think about the spec, the deck, the spec, the memo, the policy, the prompt… the medium may vary but the act of writing is how people engage meaningfully with work.
Bezos’s famous six-page memo is the most notable bet that clear writing and clear thinking are the same thing. PowerPoint, he argued, creates the illusion of understanding. Tight prose forces you to actually think and understand deeply. And even before the “knowledge work” age, writing as a mechanism held organizations together.
In the present moment, though, where the primary interface to building is now “the prompt,” every action is an editorial decision about what matters and what doesn't. Writing is no longer adjacent to work, it is the work itself. What an organization writes now gets built, immediately, at scale.
Organizations that write clearly will think clearly and build with purpose. Those that don't will write and build plenty, but they won't move with velocity or be able to tell you why. What does it mean for a business to write to the market, to its team, to its agents? And what does it hear back?
Articulation is the First Product
The friction from writing provides surface area for thinking.
Paul Graham once remarked that "Writing about something, even something you know well, usually shows you that you didn't know it as well as you thought." The struggle is the point of a first draft. You reach for a word, form a sentence, and it falls short. These moments, as frustrating as they are, show you where your understanding is thin. The difficulty of finding the right language to articulate an idea is how the idea becomes clear.
In the words of @itsbdell, “articulation is the first product.” Before anything gets built, the narrative that justifies building it has to exist. Narratives don't just communicate strategic decisions, they shape them, helping organizations combine stakeholder expectations, resolve internal contradictions, and clarify positioning against the broader landscape. The narrative is the engine.
This is the role of leadership, specifically the founder. But that function often requires a sparring partner: someone who can hold an idea in tension with you and help you hear what you're actually saying versus what you think you're saying. In practice, that person becomes something like an editor-in-chief of the ideas circulating in the organization.
Writing Runs Both Ways
Good writing has always been a conversation. When you publish, the idea gets tested against reality and comes back changed.
The best writers are good at hearing this. They don't lose the thread when reality contradicts them, they revise. The same is true of organizations. The companies that built lasting narratives did so by meticulously listening to what came back and folding it in without losing the essence of their story.
Today, AI is missing its return channel: systems where the work generates signals and those signals can find their way back to the people who need to hear them. We need a way for an organization to hear its own thinking clearly reflected in the minds of those who've heard it, a litmus test for knowing when their writing is misunderstood.
In 1494, Luca Pacioli published the first printed description of double-entry bookkeeping. Every transaction is recorded twice: debit and credit. The books have to balance. When they don't, the discrepancy is itself information. Prior to this, merchants had to be physically present to understand a business. The operation was legible only to the person living inside it. Double-entry bookkeeping was a mirror.
In the age of AI, the writing of an organization is everywhere: its prompts, its agent outputs, its customer interactions, etc. Where the writing slips and forms cracks, who hears it? And how does it get reflected back clearly enough to act on?
Building the Return Channel
Some practitioners are already building these return channels.
@emmettshine's studio Little Plains delivers what they call dual-native brand systems. The same brand is expressed in two formats. One for humans: visual identity, typography, guidelines. And one for agents: structured markdown, YAML, and JSON files that define positioning, voice, and differentiation in chunks agents can parse.
Their agents each have a soul.md. One agent, Donna, handles financial intelligence. Her file reads: "Be warm, direct, and practical. Lead with the most important insight, then support it." The file gives the agent character.
Off-brand output is a diagnostic failure not a technical one. It shows where the brand's writing is vague or contradictory. The soul.md is a hypothesis about what the brand sounds like. The agent's output is the test, and where it fails, the writing gets revised. This is a double-entry ledger for identity: what you wrote, and what came back when something tried to perform it.
@kickingkeys and @nikokozak exploration of “narrative version control” for codebases explores this thread too. Their argument was that we've stagnated because we treat the commit as the core versioning unit when the better unit is the conversation. Before code, there is always a prompt, a response, and refinement. The reasoning behind an architectural choice lives in a terminal session nobody can recover. A new engineer reads the code but not the story of the code.
Making this visible gives engineering its own ledger. Intent travels with the change and onboarding looks more like reading the story of the codebase. Others are building here too. Traces, a new product by @tarunsachdeva, surfaces the narrative trail of engineering. Entire.io, from former GitHub CEO Thomas Dohmke, is building a full reasoning trail alongside code.
Brand and engineering seem unrelated, but in both cases the work generates signal about whether the writing was any good, and someone built a system to route that signal back.
Extend this across every surface where an organization touches the world: a double-entry ledger for narrative. What gets written gets built, what gets built gets experienced, and what gets experienced talks back to help you adjust the draft.
Newsrooms, Not Software Factories
Most of what follows doesn't exist, but we believe the shape is clear enough to trace.
When an organization's working draft is legible enough for agents to carry, it becomes visible to everyone. The internal story and the external story collapse into the same document. Leadership in this new world is editing: maintaining coherence across a living text that's constantly being tested against reality.
The organization that takes this seriously starts to resemble a newsroom more than a software factory. Everyone at every level is reading, writing, and revising. New roles will emerge around signal, people responsible for noticing when the written story and the story heard out in the world have diverged, and to route that divergence to those responsible for resolving it.
Every touchpoint and interaction with users becomes a surface where the draft meets its reader. Each one will create opportunities for the organization to find out what it actually said.
The Draft is Never Final
We talk about AI in terms of what it produces, but the most consequential thing AI did to organizations was making writing the main thing they must do.
And if writing is the work, then the question that matters is whether anyone is telling you what your drafts sound like.
Build the return channel. Let the people who heard your draft tell you what they heard. The organizations that do this will compound something that most companies will never have: a clear, evolving understanding of their own story. The rest will keep writing, filling their organization with echoes of a first draft, until all they hear is noise.
