I Organized My Files and Accidentally Started Thinking 我整理了一下文件夹，然后不小心开始思考人生

This is the second in a two-part series about building a Personal OS. The first article covers the architecture and implementation. This one covers the why.

---

By late 2025, AI had crossed some invisible threshold, and after that, everything started moving at the speed of a caffeine-fueled fever dream. One week it was AI agents that would “do everything for you.” The next week, everyone was installing “skills” like they were collecting Pokémon cards. A few days later, some new open-source model arrived and the entire internet rearranged its personality around it. Social media was wall-to-wall proclamations of doom and revolution. Layoffs in news feed, sandwiched between ads for AI courses that promised to make learners “future-proof.” Somewhere, someone published a PDF called “THE 2028 GLOBAL INTELLIGENCE CRISIS” and it spread like a grease fire through social media.

I wanted to keep up. I really did. But every time I sat down to learn the latest thing, three newer things had already replaced it. I hadn’t even figured out how to write a decent prompt before the next paradigm shift told me prompts were already obsolete. The tools kept shapeshifting. The discourse kept accelerating. My brain felt like a browser with forty tabs open and every single one of them playing a different podcast at 2x speed.

So, in the middle of all this chaos, I made a decisive move. A bold, strategic play for the future: I organized my files.

---

Organizing Files Is Not Organizing Files

I should confess something: before this, my files were a disaster. My desktop looked like a digital yard sale. System files lived in the Documents folder. Project folders named “final_v2_REAL_final” sat next to screenshots from 2023. I was, in every sense, a child of the modern age. Why organize when you can search? Need something? Type three keywords and there it is. With AI, you don’t even need keywords. Just describe what you’re looking for in plain English and it finds the thing in seconds. Modern tools had made retrieval so effortless that structure felt like overhead. Like ironing your pajamas. Technically virtuous, practically pointless.

But here’s what I discovered when I finally sat down with my mess: organizing files is really organizing your own thinking. Fuzzy ideas about “what matters to me” have to become actual folders with actual names. Vague notions of “my work” and “my life” have to get untangled into real categories with real boundaries. The file system doesn’t accept hand-waving. It demands decisions.

And those decisions are not administrative. They’re cognitive.

Every time you decide where something belongs, you’re making a tiny thinking move. What is this? What does it relate to? Why does it matter? Each one is small. Over years, the compound effect is enormous. The person who makes these decisions builds a living map of their knowledge. The person who doesn’t has a search bar and a vague feeling of “I have something about that somewhere.”

One compounds. The other doesn’t.

Current education treats this like tidiness: clean up your room, sort your folders, label your notebooks. But the real skill isn’t alphabetizing. It’s classification. Relating. Structuring. Seeing what belongs together and what belongs apart. These are cognitive operations that apply to writing, to problem-solving, to communication, to project planning. And now, critically, to how well you work with AI.

So the irony of the AI age is this: as retrieval gets trivially easy, the appearance of structure becomes unnecessary. You don’t need folders when you have search. But the thinking that structure provides becomes more valuable than ever. Because search answers: Where is the thing I want? Structure answers: What do I have, and how does it relate?

The first is a retrieval problem. AI solves it brilliantly. The second is a thinking problem. And thinking problems don’t go away just because the tools got smarter. They go deeper.

---

Why Structure Matters Now More Than Ever

If organizing files is actually organizing thinking, then the next question is: why does that matter right now? Two reasons: one about the machines, one about us.

The Machine Side: Structure Is Bandwidth

Here’s where it gets concrete. People who can give AI structured context about their thinking, goals, projects and knowledge get exponentially more value from every interaction. Those who can’t are stuck at the “hey AI, find my file” level.

Structure is the bandwidth of the human-AI interface.

When I sit down with AI and it already knows my mission, my active projects across six life domains, my thinking frameworks and the patterns connecting my learning to my writing to my working, the conversation starts at a completely different altitude. I’m not explaining context. I’m extending thought.

And this gap will only widen. AI’s evolution is heading in two clear directions: connecting to the outside world through protocols like MCP, and connecting to your personal digital space through local system access. Both paths lead to the same place: AI that knows your stuff and can act on it. The trajectory points toward personal agents that anticipate, initiate and synthesize on their own. Personalized models fine-tuned on your data and preferences. This isn’t speculative. It’s the visible roadmap.

When that future arrives, people who have a clearly structured representation of themselves (their knowledge, their goals, their thinking patterns, their values) will be able to plug right in. People who have a scattered pile of files and a search bar will be starting from zero. Structure isn’t just useful today. It’s infrastructure for a future that’s arriving faster than most people expect.

The Human Side: Agency Is Architecture

But here’s the part that matters more. No matter how capable AI becomes, its ultimate purpose is still to serve human needs. And human needs come in two kinds. There are the general ones (communication, productivity, information) that platforms and apps can define for millions of people at once. Then there are the specific ones. Your specific ones. What you’re trying to build with your career. What kind of person you want to be. Which skills matter to your particular path. What trade-offs you’re willing to make right now, and which ones you’re not.

No platform can define those for you. No AI can figure them out on your behalf. And yet, here’s the problem: most of us haven’t defined them clearly for ourselves either.

So what happens when a person who doesn’t know what they specifically need meets a machine that’s endlessly capable of doing things? They delegate. AI agent platforms are going viral. People are downloading skills, automating workflows, handing off task after task. The results are impressive: things get done faster, problems get solved more efficiently. But there’s a quiet cost. When you delegate a task, you gain time but lose the cognitive work that task involved. Some of that work is mechanical, and good to offload. But some of it was building your understanding, forcing your decisions, creating the micro-learning that accumulates into expertise. AI can search, summarize and synthesize in seconds. But it cannot learn for you. It can bring you the materials. It cannot build the house in your mind.

And without a clear sense of what you need, the delegation doesn’t just save you time. It starts steering you. Which ideas AI surfaces, which patterns it reinforces, which directions it nudges. All of it quietly reshapes how you think. Not because AI has an agenda, but because tools shape their users whether anyone intends it or not. The person who automates everything without knowing what they want doesn’t become more productive. They become more dependent, moved by invisible strings they never chose.

This is why building structure yourself matters. Not for efficiency. For clarity.

When I sat down to write my identity file (who am I, what do I value, how do I think, where are my blind spots), I was doing something AI cannot do for me. I was defining my own needs. Not in the abstract journal-prompt way. In the structural way: what are the actual categories of my life? How do they connect? Where is energy flowing and where is it stuck?

Every architectural decision forced a thinking decision. Should “parenting” be separate from “family”? That’s not a folder question. It’s an identity question. Should “learning” sit inside “work” or stand alone? That depends on whether I see learning as instrumental or intrinsic. The answers revealed things about my values that I’d never fully articulated until the structure demanded it.

The person who does this work knows where they are and where they want to go. Like someone who writes a clear brief for their own role, not a job description imposed from above, but one they author themselves. It becomes an expectation they set. And then, every interaction with AI becomes a calibration: Am I moving closer to what I actually want, or drifting further from it?

That’s agency. Not the freedom to do anything, but the clarity to know which things are worth doing.

---

How I Built It

So I did what felt right in the middle of all the chaos: I organized my files into what I call a Personal OS. It’s a small set of structured markdown files, roughly 20 to 40, that represent my self-model. Who I am, where I’m going, how I think, what I’m working on and how it all connects. AI reads this first in every session. The result isn’t just convenience. It’s a fundamentally different kind of collaboration.

What Makes This Different

The concept of a “personal OS” isn’t new. People have talked about digital twins, personal knowledge management, second brains, personal language models. There’s a growing community of hands-on practitioners building sophisticated systems (tools like Tiago Forte’s Building a Second Brain, August Bradley’s PPV in Notion, or the Obsidian-based Zettelkasten revivalists) and much of that work is genuinely impressive.

But most of these projects share an implicit assumption: the human configures the system, the system serves the human. The arrow goes one way.

Every time you interact with AI, two things happen at once. You shape the AI’s behavior through your prompts, your feedback, your configuration choices. And the AI shapes you, through which ideas it surfaces, which patterns it reinforces, which directions it nudges your thinking. The question isn’t whether mutual construction is happening. It’s whether you’re aware of it and doing it intentionally.

Think about what happens when a person with no clear self-model starts using AI heavily. They ask AI to organize their priorities. AI suggests a structure. They accept it. They ask AI for a career plan. AI generates one. Sounds reasonable, so they follow it. Over time, AI is making more and more of the framing decisions (what to focus on, what to deprioritize, how to categorize their own life) and the person’s big-picture thinking quietly atrophies. They’re still in the driver’s seat technically, but the GPS is choosing all the routes. That’s what I mean by invisible strings.

Now think about the opposite. A person who has already done the work of defining who they are, what they want and what their current reality looks like. When they hand AI their self-model, the relationship flips. It’s like the difference between a new hire with no job description and one who wrote their own role brief. The first drifts toward whatever tasks land on their desk. The second has an expectation, a clear picture of where they are versus where they want to be, and every interaction becomes a calibration. Is this moving me closer, or further away?

That awareness changes everything, including the design of the system itself. The system I ended up with has three layers: a “brain” layer (small, structured, the meta knowledge about “me”), a “body” layer (where actual life and work data lives, selectively shared with AI), and a private vault (encrypted, AI-excluded). That three-layer split maps to something permanent about being human: the public self, the working self and the protected self. The exercise of drawing those boundaries (deciding what AI should see, what it can access on request, and what it should never touch) is itself a practice in self-knowledge.

Where the Compound Interest Lives

The most valuable form of intelligence isn’t depth in a single domain. It’s the ability to see that a pattern in one domain applies to another. The systems thinking you apply at work can reshape how you approach writing. The writing insight you had last Tuesday connects to the product architecture you’re stuck on. The architecture you developed for yourself turns out to be exactly what your learning framework needs.

This kind of transfer is almost impossible in a flat, unsorted information landscape. You only discover connections between things when those things have some kind of spatial or categorical relationship, when they live in a structure that lets you see across domains rather than just within them.

My Personal OS was explicitly designed for this. The “connections” layer (a living map of active work across domains, plus AI-maintained cross-links and a transfer log) exists to make visible what would otherwise stay hidden. With some settings, AI can help me check periodically: does anything from this conversation connect to another active domain? A messy desktop is a collection of isolated dots. A structured system is a network. And networks generate emergent properties that their individual nodes never could.

---

Starting Point, Not Destination

My Personal OS is not finished. It won’t ever be finished. It’s a living system. The identity file will be rewritten as I change. Goals will shift. New frameworks will emerge. Old domains will archive and new ones will appear. And because the whole thing is version-controlled with git, every evolution is recorded. The changelog becomes a human-readable version history of me, accumulating entries over years into a compressed narrative of personal evolution. The diffs between versions will show how my thinking changed, not just what I thought.

The system evolves on three fronts at once. The self-knowledge deepens:

• I get better at articulating who I am and what I’m building.
• The tools mature: AI becomes more capable of proactive synthesis, working across my domains without being asked.
• And the practice itself sharpens: I learn when to give AI what level of context, what to share and what to keep private, how to be more intentional about what goes in and what stays out.

The value isn’t in having a perfect system. The value is in the ongoing practice of structuring, examining and articulating who you are and what you’re building. That practice keeps you in the driver’s seat during a stretch of history when it’s very easy to become a passenger.

---

The Only Question That Hasn’t Changed

The world is changing faster than any of us can track. New AI models drop monthly. Yesterday’s workflow is today’s legacy. The tool you just learned is already being replaced by a tool that learns for you. If you feel a low hum of unease underneath all the productivity gains, you’re not imagining it. Here’s what I think that feeling is really about: we’re losing the sense that we know what we’re doing and why. The tools keep changing, but the deeper question hasn’t changed in thousands of years. Who am I, and what am I trying to build with my time?

He who has a why to live can bear almost any how.

— Nietzsche

The “how” changes every month now. The “why” is what keeps you from drowning in it.

And maybe that’s the real discovery hiding inside this whole exercise. I set out to organize my files. I ended up confronting the same question that Socrates badgered people about in the Athenian marketplace, that Montaigne circled in his tower, that every contemplative tradition eventually arrives at: know yourself. The tools for asking that question have changed beyond recognition. The question itself hasn’t moved an inch.

A Personal OS won’t answer it for you. But it will force you to keep asking. And in an age where everything else is being automated, the willingness to sit with a hard question rather than delegate it might be the most human thing left.