More of us is possible
There is a researcher in the person who keeps asking why. A writer in the person who wants to explain what they found. A builder in the person who wonders whether the idea could become something real. Often, these are the same person. The calendar is less accommodating.
We learn to put parts of ourselves on hold. The research waits while we build. The writing waits while we keep the business running. An interesting question survives in a folder until returning to it requires more time than we can find. Eventually, we mistake the difficulty of resuming for a lack of interest.
In Limitless, Without the Pill, I wrote about building Orionfold across research, software, publishing, design, and operations. What moved me was the possibility of bringing different modes of thought into the same working life. Context preserved between them made those transitions more practical.
That experience left me with a question: how much of our ambition is constrained by the cost of keeping it alive?
The weekly brief, the editorial review, the plan for a project we care about: each needs somebody to return, check the sources, update the numbers, and recover the reasoning. The cost is small enough to overlook and persistent enough to shape what we attempt.
Living Documents begin here. A page can preserve its sources, carry a bounded routine, and leave evidence of the work done on its behalf. The purpose is to help knowledge remain useful while its owner turns to another part of life. A little less reconstruction can create room for a much larger question.
Give curiosity a place to return
A notebook is a generous invention. An observation, a diagram, a calculation, and an unfinished thought can share a page. It allows a person to follow connections before deciding which discipline they belong to.
The modern working document can offer that same generosity. An argument can sit beside its source table, a chart, a diagram, and an image. In Flow, these views grow from readable content in the file. Editing a cell or changing a chart can remain part of working on the page. The documents and files tour explains that foundation.
But a useful page also contains an invisible half. Somebody knows which sources matter, how the numbers fit together, what deserves attention, and which conclusion requires a decision. When the page travels without that method, its next reader inherits a reconstruction job. A month later, that reader may be its author.
A Living Document gives the method a home. In Flow, a document can name files, folders, and public web sources, collect and shape data, and bind its charts and tables to the resulting files. Standing jobs describe what should happen again. The updated page gives a person something to understand and revise.
This makes an old question newly useful: what would let me return to this work with my curiosity intact? The answer might be a refreshed table, a source check, or a clear account of why an update could not run. Flow provides editors for jobs and data definitions, including previews. A local model can draft a definition from a sentence for you to review. The method remains yours to author or accept.
A better kind of morning
Imagine a creator who is also building a small business. She keeps an operating review beside a folder of sales exports, campaign results, and a few public sources worth watching. The document holds measures she has chosen and questions those measures may help her answer.
During the day, she can ask Flow’s Agency to proofread, summarize, translate, or develop a document. Agency proposes exact changes for her review before saving. A standing Night Shift job makes a different agreement: she declares bounded maintenance in advance, then reviews its versioned results afterward. These are distinct permissions, explained in writing with AI.
During the next Night Shift, Flow collects the declared data and redraws the bound tables and charts. With a paid plan and a local model configured, Overnight notes can add a few sentences about the page’s numbers and what moved.
In the morning, she opens the changed document. Updated parts are marked in place, with receipts available to inspect. These are changes already applied under the standing job’s authority. She can keep or revert each part.
Perhaps a campaign found an audience. Perhaps a missing source prevented an update. Perhaps the numbers barely moved. Each outcome should help her see where the work stands. A quiet night may leave only two useful sentences.
Now she can bring a different part of herself to the page. The analyst notices a pattern. The writer finds a question in it. The builder considers an experiment. The routine prepared the material; the connections belong to her. That is the morning I want Flow to make possible.
The document can carry the responsibility.
The decision still belongs to you.
Keep the right to question the result
The appeal of capable AI can make verification feel like an interruption. Yet the person whose name is on the work still needs to understand what they are accepting. A plausible paragraph can take longer to verify than it took to generate. A beautiful chart can conceal a poor comparison.
In Towards a Science of AI Agent Reliability, researchers evaluated 15 models across two benchmarks. They found only small improvements in reliability alongside recent capability gains, and distinguished successful answers from consistent behavior, robustness, and predictable failure. Those results describe the evaluated models and tasks; they are no measurement of Flow. Rabanser and colleagues
Flow’s Overnight notes begin with a deliberately narrow assignment. The model receives the page’s own numbers. Every number in its draft is checked against a source row. If a number cannot be traced, Flow asks for a correction and withholds the notes if the problem remains.
This establishes that a number has a source in the supplied data. It cannot establish that the source is correct, the interpretation is fair, or the conclusion deserves publication. Your experience and judgment still matter at each of those steps.
Flow’s receipts and evidence keep a record beside the work: the model and route, cost and its basis, checks and their results, and revisions before and after a change. An estimate must remain distinguishable from an observed charge. A previous version provides a way back.
A person can take a larger creative risk when a smaller operational mistake is easier to see and recover from. Visible evidence gives exploration a firmer footing.
A larger life needs quieter tools
Gallup’s July 2026 workplace report found that, among US employees using AI, 51% used it for writing and editing, 49% for search or research, and 16% for automation. These are reported uses, not measured productivity gains. They show how familiar the first forms of assistance have become. Gallup
I expect recurring work to become more accessible too. The question is what it will ask of us. Twenty overnight updates can become twenty new obligations. A tool that keeps finding work for its owner may leave the day feeling smaller.
The unit of progress should be a responsibility maintained well. Can I return to the research brief without reconstructing it? Is the operating review easier to question? Has the method earned enough confidence that I spend less time supervising each routine?
Sometimes the best result is no change. A source check finds nothing new. A draft is withheld because a number could not be traced. A job stops and explains what it needs. Restraint belongs inside the definition of useful work.
Flow’s Ideas board brings possible next steps beside the document; the person chooses what to pursue. Agency carries the chosen work through scope, route, review, and a recorded result. Suggestions help when they make the next worthwhile decision easier.
The attention returned need not become another task. It can become reading outside your field, a conversation, a walk, or time with somebody you love. A broader working life should leave room for the rest of life too.
Own the work and its boundaries
A file is a simple kind of independence. You can keep it, copy it, and return to it later. As a document begins doing work, ownership must also cover the method, the inputs, and the authority to act.
Flow starts with Markdown files in folders you choose, without importing them into a private library. Jobs and data definitions stay with the work; receipts preserve its record. Another editor will not run Flow’s jobs, but it can read the words. Search and links help you find your way back through your own material.
Night Shift runs on the Mac. Its helper can work with Flow closed while the Mac is available, idle, and plugged in during the window you choose. If the Mac sleeps through the window, Flow can catch up once after it wakes. A powered-off computer cannot complete local work.
Daytime model work offers an explicit choice of place. Flow includes a local runtime, can use compatible models already in Ollama or LM Studio, and supports hosted providers you choose. The route should explain where the text goes and what the run is expected to cost. Permission to use a local model does not authorize sending the document elsewhere. The models and runtime tour explains these boundaries.
Night Shift’s model work stays on the Mac, without quietly substituting a hosted model. Declared public web sources still require network requests. A useful routine should make these ordinary conditions understandable.
Ownership extends to the business model. Flow Base keeps reading, writing, searching, organizing, charts, diagrams, and exporting free, along with deterministic Night Shift maintenance. Model-powered actions belong to Pro, including those using a local model. If a subscription ends, the documents remain open and editable, and existing receipts remain readable. Your knowledge should remain available as your circumstances change.
FOLD: turn one intention into a practice
The bridge between a large ambition and a useful document is a small practice. For Living Documents, I use an adaptation of FOLD: Frame the outcome, Organize the routine, Leave a receipt, Develop the practice.
Frame the outcome. Choose a page you already reopen. Write down what would make returning easier: a weekly view of campaign results, a research brief with current sources, or a budget whose chart follows new data. A clear outcome helps you recognize a useful result and reject a distracting one.
Organize the routine. Name the sources, the permitted changes, and the rhythm. A researcher may watch selected public pages. A creator may work from their own exports. A small team may begin with files it already exchanges. Flow’s editors let you define the recurring work and inspect its definitions. Begin with a task small enough to understand completely.
Leave a receipt. Run a cycle and inspect what happened. Read the changed part alongside its sources and checks. Notice what failed and what stayed unchanged. Keep or revert the result. A receipt helps the future you understand the choice the present you made.
Develop the practice. Return for a second cycle. Correct the definition that was too broad. Remove the update that added noise. Keep the comparison that helped you see something. A method becomes valuable through the care accumulated in it.
Flow includes seven Living Document examples as starting points. Their sources and definitions are available to explore and adapt. None can choose your purpose for you. FOLD is a way to make that purpose durable enough to test, then carry what you learn into the next cycle.
Let hard-won knowledge travel
An experienced researcher knows which distinctions a beginner may miss. An editor knows which questions strengthen a brief. A thoughtful operator knows which measures belong together and which comparison would mislead. Expertise lives partly in those choices.
A book can explain them. A template can preserve their shape. A Living Document could carry some of that expertise as a practice: sources, definitions, examples, and review criteria that a reader can inspect and adapt while doing real work.
Some is an essential qualification. Judgment cannot be fully packaged. A useful method reveals its assumptions and makes clear where interpretation begins. It leaves room for the learner’s experience to change it. Its value may lie as much in the mistakes it anticipates as the steps it performs.
This is a future I want to explore: richer domain methods and documents that cooperate through declared relationships. These are future possibilities, not a claim about Flow’s current feature set. They should grow from individual documents that people already find useful. A change in one brief must not silently rewrite conclusions elsewhere.
There is a personal possibility here as well. The researcher’s care can improve the writer’s argument. The builder’s experiment can give the researcher a better question. A method preserved between them gives curiosity more chances to compound.
The lasting asset is the practice a person has refined, together with the work and reasons they chose to keep. As models improve, that asset should become more useful and remain something its owner can carry forward.
The document has to earn the return
The larger software platforms are pursuing recurring work too. Microsoft describes Copilot Cowork as handling complex, long-running tasks across tools. Many people will sensibly choose capabilities already available in their workplace. A new product has to make a specific contribution. Microsoft
Flow brings standing work into readable files you own, runs Night Shift on your Mac, and returns you to marked, reversible changes in the document itself. That combination has to prove useful on a real responsibility. A familiar page is an invitation to try; its continuing usefulness is the reason to stay.
It will not fit every task. Some work depends on systems the document cannot access. Some changes require a conversation before any update makes sense. Sometimes a calendar reminder and a few minutes of direct work are sufficient.
Even a well-chosen Living Document can begin with poor sources or a mistaken definition. The method will need editing. The question for the second cycle is whether the work becomes easier to maintain and the result easier to understand.
The studies in this essay provide context for that design challenge. They do not establish that Flow improves productivity, accuracy, or creativity. Those are outcomes we must examine through actual use. The person returning to their own document is the one who can judge whether the routine has earned its place.
Make room for the next chapter
The Renaissance idea that appeals to me is a person whose interests can inform one another. A builder who reads widely. A researcher who makes things. A writer who follows a question far enough to learn a new discipline. Depth and curiosity can grow together when the cost of moving between them becomes manageable.
Living Documents offer one practical piece of that possibility. You close a page without abandoning its continuing responsibility. You pursue a new question without letting every previous thread become expensive to recover. When you return, the page shows what changed, what held, and what still needs you.
Some work should remain finished. A poem can keep its last line. A letter can belong to the day it was written. A Living Document earns its name through the responsibility it serves. We should be as free to end a routine as to begin one.
Choose one document you already reopen. Give it a purpose, a small job, and a result you can judge. Let it complete a cycle. Keep what helps. Revise what does not. The Living Documents Manifesto is the short expression of this conviction; the practice begins in a file of your own.
I want the knowledge we have worked to acquire to keep helping us while we discover what else we might become. The direction remains ours. Our tools can give more of our curiosity a place to go, and more of our lives room to unfold.