Summary: The article uses the author’s 30 years managing type 1 diabetes as a lens for a broader framework called “legibility”—how much context outsiders need to understand something, and who controls that translation. Drawing on James C. Scott’s Seeing Like a State, it argues legibility now operates across four scales: the self, public identity, organizations, and machines. The core thesis for business audiences: organizations fail not from bad strategy but from becoming illegible to themselves—when culture, story, and systems tell different stories, and AI systems amplify whichever version they can read — often the wrong one. Wikipedia’s 2026 ban on LLM-written articles is used as a case study of a community with enough internal clarity to catch its own tools drifting from its values.

What thirty years of making my body readable taught me about building companies and communities in the age of intelligence.

How legible are you? To yourself? To your family? To your colleagues, your company, your insurance, or your government? 

Legibility is never neutral. The way you make yourself legible determines what others can understand, measure, reward, deny, or misunderstand about you. In the age of intelligence, the work is to design your legibility: to decide what becomes readable, to whom, through which interface, and at what level of compression. When you exercise agency over your legibility, you keep your humanness at the wheel instead of allowing institutions, audiences, or machines to define you by default. 

I have spent most of my life trying to make myself legible to machines, doctors, insurance companies, and most of all to myself. I was diagnosed with type 1 diabetes in 1995, when I was twenty-one months old. At the time I used large needles to inject insulin and pricked my finger when I needed to check my blood sugar.

My body cannot regulate its own blood sugar, so I do it manually. Every day! Too high and I get lethargic, too low and the world begins to sink away. Managing type 1 diabetes is kind of like flying a plane. I’m always making small corrections, always checking alongside the instruments, always trying to keep the needle flying in a safe altitude while I live a normal life and eat ordinary food.

To fly that plane I have to make my glucose legible. First to myself, with a meter or a monitor. Then to my doctor, who reviews my numbers every few months and runs a test called an A1C that compresses ninety days of my life into a single average. My doctor makes me legible to my insurance company, translating my numbers into a request for the equipment I need. Each step is a kind of translation. Each translation strips away context and leaves a cleaner, smaller version of my truth.

I want to talk about that word, legible, because over the past few months it has become a lens I check almost everything through. Legibility is how much context you need to understand something. A highly legible thing can be read by an outsider with little background. An illegible thing requires deep, local knowledge to interpret. 

The tension is that legibility creates both coordination and vulnerability. It allows a doctor to treat you, a colleague to work with you, or a machine to process alongside you. But every act of legibility also compresses context. It produces a cleaner, smaller version of the truth.

This article follows that tension across four scales: the self, the public identity, the organization, and the machine. At each scale, the questions are: who gets to read you, what do they lose in the translation, and how much control do you retain over the version they see?

Seeing like a state

The idea of legibility comes from James C. Scott’s Seeing Like a State. His argument is that pre-modern people lived in high-context worlds. They knew each other deeply, used different names in different settings, and lived in cities that sprawled like organisms. None of this served a king who needed to count and tax his subjects. So states imposed legibility from above. They handed out fixed surnames, put cities on grids, mapped and registered land, and standardized the world into something readable from the center. It served the people doing the reading (or tax collecting) far more than the people being read.

Scott’s warning is that this flattening does damage. The local knowledge and improvised skill he calls metisgets crushed under the administrative grid. And so, throughout history, people have made themselves deliberately harder to read as a form of protection. They returned to oral culture, moved from farms into forests, and did whatever made them harder to tax and conscript. Illegibility has always been a strategy for staying free.

That instinct never went away, and it migrated online. The internet has its own version of the state: the algorithm, the mob, the careerist, the collapse of context that happens when everyone reads the same post at once. A scene that becomes too legible gets gamed. People who can read the status signals but do not share the community’s values move in, and the original builders move out. This is an old story of how subculture gets discovered. What I keep circling back to, however, is who holds the power now.

The digital age handed us partial agency

When Scott wrote about legibility, it was something done to you. The state read you whether you liked it or not. The digital age adds something new into the equation: partial agency. Individuals can decide what to publish, what to withhold, how to describe themselves, and which version of themselves they intentionally present. But platforms still rank, infer, index, and recombine those signals in ways the individual does not fully control. 

I have been watching this play out in the type 1 diabetes community online. People take the most private, frustrating data of their lives and decide how to make it legible to others. Some lean all the way into the hard parts and build a following around how brutal the disease can be. I do not judge that for a second, because the disease is brutal. Others build a following around posting their flawless numbers and clean graphs. Between those poles sits a wide spectrum, and every account is really a series of choices about how legible to be, on which topic, to whom. Each person is deciding, day by day, how much of their context to compress into something a stranger can read.

Something became obvious fast: visibility and legibility are different axes. An account with a million followers can be far more legible than another account with the same numbers. Reach is a different thing from being understood. For anyone building a brand, a company, or a public self, that distinction is crucial to understand. The goal is to be read clearly by the right people. Going viral does not deliver that on its own.

Read yourself first

Here is the move that changed how I work. The most valuable use of legibility is internal rather than external. You have to make yourself legible to yourself before you can be legible to anyone else.

The clearest example is the date-me document, a practice where people write up, in detail, who they are and what they are looking for in a partner. The obvious value is filtering inbound interest. The real value is the hours you spend forcing fuzzy wants, desires, and feelings about yourself into clear language. The document ends up being the residue of that practice of thinking. You cannot communicate clearly to the world what you have not yet made clear to yourself. Writing it down on a date-me document with the intent for others to see is one great way to figure out how legible you are to yourself.

This is why I tell the founders and builders I work with to write the thing down before they pitch it to anyone. Write what your company is for. Write what you will not do. Write what you actually want out of this, in words precise enough that you would recognize the right opportunity when it landed in your lap or inbox. That kind of precision work is inward legibility, and it is the cheapest, highest-leverage work available to you.

For me, the glucose number does a similar job. It is my current state reflected back at me, in a form I cannot argue with. That is also why making myself legible is hard. When I make a mistake, it shows up as a number, and those numbers can accumulate into fear, frustration, and distraction if I let them. Reading yourself honestly is uncomfortable because there is nowhere to hide. That discomfort is the point. It is the same discomfort a company feels the first time it looks honestly at what its culture actually rewards, rather than what it claims to value.

Self-legibility is the foundation of every other form. A person, company, or community cannot deliberately manage how others read until it has developed an honest account of itself.

Design the front door

Once you can read yourself clearly, you can begin designing how others encounter you. You begin to ask: legible to whom, about what, on which channel, at what level of compression?

The pattern that resolves most of the tension is a clear, legible front door, with as much complexity as you want once people are inside. The pitch “we are Uber for cats” is pure front-door legibility. It lets an investor slot you into a model they already hold and move the conversation forward. The nuance and the real differentiation come later, after the simple first impression has done its job of getting you in the room.

Consider a small cashmere company with exceptional products but an unclear message and stagnant growth. Instead of chasing large, highly visible influencers to help push their products, they instead partner with a network of micro-influencers who are deeply legible to a specific audience: people who already understand the value of high-end materials, provenance, and craftsmanship. These influencers may not be widely known, but within the right circles, their recommendations are trusted and made legible to the demographic that wants high-end cashmere. The company designs its front with a simple line: “premium, traceable cashmere from a single Mongolian region”. This ensures that both its messaging and its partners reinforce that clarity. Sales begin to climb because their message is made legible to the right audience who understands it immediately. Legibility to the right audience compounds faster than visibility to the wrong one. You are not legible until your customers can explain you accurately when you’re not in the room.

And legibility flows in many directions inside any organization at once. Employees making themselves readable upward. Leaders making the vision readable downward. The company explaining itself to customers, customers explaining it to each other, the company explaining itself to investors. Each of those is a separate translation problem with its own right level of compression.

Organizational legibility is the degree to which culture, story, and systems describe the same organization. A company may tell a clear story while rewarding behaviors that contradict it. It may build technically sophisticated systems around assumptions its employees do not share. It may have a healthy internal culture that customers and investors cannot understand from the outside. In each case, one layer is legible while the larger organization remains only partially legible. 

This is why legibility must move across the whole system. Culture reveals what people believe and reward. Story makes those beliefs coherent and communicable. Systems turn them into repeated decisions, workflows, and consequences. When those layers reinforce one another, an organization becomes readable to itself and trustworthy to others.

Selective legibility

The most sophisticated move is to be fully readable to exactly the right people while staying opaque to everyone else. The mechanism is context as a filter. A word like optimistic means nothing threatening to a casual reader. To the right reader it signals an entire worldview, a builder who believes the future can be better and is willing to work on it. The same word filters one audience out and pulls another in, and it does the filtering without ever sounding hostile.

This is how small creative communities stay coherent while resisting invasion. Insider vocabulary, shared references, aesthetic choices that only land if you already have the context. They exclude by requiring context, not by being cruel. When the signals become too widely understood and stop filtering anything, the culture cannot resist invasion. Then the community has to retreat into new illegibility, find fresh signals, and rebuild its context, or accept that the scene has changed into something else. For a founder, the practical version is this: design your public signals so the people you most want to reach feel an immediate jolt of recognition, while the people who would only add noise simply pass by without friction.

Legibility to machines

There is a new reader in the room now, and it complicates the stakes of legibility further. We are starting to make ourselves legible to machines.

You can watch it happening in real time. People are writing llms.txt files so language models can read their sites. They are building machine-readable interfaces, experimenting with protocols such as MCP, and reformatting documentation into clean markdown so AI tools can parse it more reliably. We are restructuring our interfaces to accommodate the reading habits of software that is supposed to be serving us. The irony is hard to miss. We are bending ourselves to be readable by systems we built to read us.

The pattern resembles the process Scott described: institutions restructure complex human realities into forms their systems can read. But there is an important difference. There is technically no single king this time. There is a new kind of reader with new constraints, and the same old question underneath: how much context do you need to understand something, and who gets to decide what context is even available? In the best case, legibility to humans and legibility to machines converge, because a clear, self-documenting interface serves both. Right now they often diverge, and closing that gap is one of the central design challenges of the intelligence age.

I see this as the next chapter of the legibility story, and it is the one most relevant to anyone building today. My core belief is simple. In the age of intelligence, we have to develop dynamic ways to manage our own legibility: to ourselves, to each other, to our companies, and now to the models that read all of it. Brands and companies can be made more or less legible to these systems on purpose, the same way I decide how legible my glucose data is to my doctor. The people who do this deliberately will be understood. The people who let it happen by default will be read in ways they never chose.

The same force at community scale

Wikipedia offers a test of this argument at a community scale. What happens when a community makes its principles explicit enough to judge the behavior of its automated systems? 

In March 2026, the English Wikipedia community voted 44 to 2 to ban large language models from writing or rewriting article content. It’s easy to read that as a community flinching at a new technology, but it’s something more interesting. It was a community that had spent twenty years learning what its own automated tools were doing to it, and that finally had the cultural clarity to act. Years earlier, Wikipedia’s quality-control bots, built to protect the encyclopedia, had quietly driven the survival rate of good-faith newcomers down to around twelve percent. The bots learned from what the community actually did rather than from what it said it valued, and then they ran those patterns at machine speed. The community drifted away from its own ideals one automated decision at a time.

Wikipedia could eventually correct course because it had made itself legible to itself. Its core principles were named, debated, and written down clearly enough that members could recognize when their tools were betraying those principles. Most companies and communities do not have that. They have a culture they have never made readable, so when AI tools start mediating their work, they have no clear self to measure the drift against. That, in one sentence, is the whole imperative. Make your culture legible to yourself before the machines read it for you and quietly decide who you are.

What I am taking forward

Legibility works best understood as a means rather than a goal, and it comes with known trade-offs and known countermeasures. You are a person of free will deciding how legible to make yourself.

After thirty years of turning my body into numbers, here is where I have landed. The work is to design your legibility rather than to maximize or minimize it. Decide what you make readable, to whom, at what compression, through which interface. Read yourself clearly first. Build a legible front door. Protect the complexity behind it. Stay aware of every new reader who shows up, including the ones made of software. Know which parts of yourself you are willing to compress, and hold the line on the parts you are not.

We increasingly find that organizations fail because they become illegible to themselves. Leaders may say one thing, incentives reward another, customers experience a third, and AI learns from and expands upon a fourth. Our work is helping organizations recover enough legibility that they can align culture, story, and systems before intelligent machines amplify the wrong version of who they are.

REFERENCES

Scott, J. C. (1998). Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press. Full text (PDF)

swyx, Krouse, S., & Karlsson, H. (2026). Legibility Explained (by People Who Don’t Hate Legibility)[Video]. YouTube. https://www.youtube.com/watch?v=96S_64ipHOA

Price, L., & Cubeta, C. (2026, May). From Visibility to Legibility: Rethinking Content in the Age of AI Discovery [Video]. The Foundry, AgenticAdvertising.Org. YouTube. https://www.youtube.com/watch?v=l61df4mggOY

Wikipedia: Writing articles with large language models/RfC. (2026, March). English Wikipedia. Retrieved May 2026. https://en.wikipedia.org/wiki/Wikipedia:Writing_articles_with_large_language_models/RfC

Halfaker, A., Geiger, R. S., Morgan, J. T., & Riedl, J. (2013). The rise and decline of an open collaboration system: How Wikipedia’s reaction to popularity is causing its decline. American Behavioral Scientist, 57(5), 664–688.

Howard, J. / Answer.AI. The llms.txt proposal. https://llmstxt.org

Anthropic. Model Context Protocol (MCP). https://modelcontextprotocol.io