Your best ideas are not the ones you have now
They are the ones you will have after revising the ones you have now. And after revising those. And after revising those again.
Darwin opened his first notebook on the transmutation of species in July 1837. He sketched a genealogical tree, scrawled "I think" above it, and began working through the implications of species changing into one another. He did not publish On the Origin of Species until 1859 — more than twenty years later. During those two decades, his theory moved through Notebooks B, C, D, and E, each one refining, contradicting, and extending the previous. Natural selection did not appear as a concept until Notebooks D and E in late 1838 and early 1839. The version of the theory the world eventually read was not a single insight. It was the product of sustained, tracked revision.
Most people do not work this way. They have a thought, they write it down (if they write it down at all), and when their thinking changes, they either forget the old version or overwrite it. The result is a knowledge system with no depth — only the latest snapshot, with no record of how it got there.
Versioning your atoms means keeping the history. Not because old versions are always right, but because the trajectory of revision is itself knowledge.
What software engineers already know
In April 2005, Linus Torvalds built Git in roughly ten days. He needed a version control system for the Linux kernel — a codebase with millions of lines of code and thousands of contributors — and nothing available met his requirements. He wanted something fast, distributed, and designed to prevent data corruption.
The core insight behind Git, and behind every version control system before it, is that the history of changes is as valuable as the current state. A codebase is not just the code that exists right now. It is every change ever made, who made it, when, and why. You can revert to any previous state. You can see exactly when a bug was introduced. You can compare two versions side by side and understand what shifted.
Git now dominates software development — nearly 95% of developers use it as their primary version control system. The reason is not nostalgia for old code. It is that understanding how something evolved tells you things that the current state alone cannot. A function that has been rewritten eight times signals architectural uncertainty. A module that has not changed in three years signals stability. The diff — the record of what changed between versions — is often more informative than either version on its own.
Your ideas deserve the same treatment. When you revise a belief about leadership, about how markets work, about what makes a good relationship — the revision itself carries information. The pattern of your revisions carries even more.
The philosophy of rational belief change
Epistemologists have formal frameworks for this. In 1985, Carlos Alchourrón, Peter Gärdenfors, and David Makinson published their landmark paper on the logic of theory change — now known as AGM theory. It defines three operations a rational agent performs on beliefs:
- Expansion: adding a new belief without checking whether it conflicts with existing ones.
- Revision: adding a new belief while maintaining consistency — which means something else has to give.
- Contraction: removing a belief entirely.
The AGM framework is not a productivity hack. It is a formal model of what it means to change your mind rationally. And the key insight is that belief revision is not a single event — it is an operation with inputs, outputs, and constraints. When you revise a belief, you are not just replacing one thought with another. You are navigating a web of commitments: what else must change to accommodate the new belief? What evidence triggered the revision? What previous beliefs were contracted to make room?
Without version history, you cannot answer any of these questions. You can tell me what you believe now. You cannot tell me what you gave up to believe it, or what evidence would make you give up your current position. Versioning your atoms is what makes the difference between holding beliefs and managing beliefs.
Scientists never erase — they cross out and initial
Laboratory notebooks are one of the oldest versioning systems in human practice. The NIH's guidelines for lab notebooks make the protocol explicit: write in permanent ink, never tear out pages, never erase. If you make an error, draw a single line through it so the original text remains legible, initial the correction, and date it.
This is not bureaucratic overhead. It is epistemically fundamental. A scientist who erases a wrong hypothesis destroys evidence of the reasoning process that led to the wrong hypothesis — and that reasoning process may contain clues about what the right hypothesis is. The history of what you tried and why it failed is part of the knowledge, not noise to be cleaned up.
Marie Curie's lab notebooks are still radioactive more than a century later. They are also still readable. Every measurement, every calculation, every crossed-out line tells the story of how two Nobel Prizes were earned — not through sudden inspiration, but through tracked, methodical revision of what was known.
Your personal knowledge system should work the same way. When you change your mind about something — when a management principle fails in practice, when a mental model breaks under new evidence, when a relationship dynamic you thought was fixed turns out to be more nuanced — do not delete the old version. Cross it out and initial it. Metaphorically or literally.
The Ship of Theseus problem — and why it matters for your ideas
The ancient Greeks posed the question: if you replace every plank of a ship one at a time, is it still the same ship? Thomas Hobbes added a sharper version: what if you build a second ship entirely from the removed planks? Which one is the Ship of Theseus?
Your ideas face the same problem. Your understanding of "leadership" has probably changed every component since you first encountered the concept. Is it still the same idea? Your belief about work-life balance in your twenties shares almost no content with your belief about it now. Same label, completely different substance.
Wittgenstein lived this problem publicly. The Tractatus Logico-Philosophicus (1921) argued that language maps directly onto the logical structure of reality, and that philosophical problems arise from misunderstanding this mapping. The Philosophical Investigations (published posthumously in 1953) rejected nearly every core claim of the Tractatus, arguing instead that meaning comes from use within "language games" — not from logical correspondence. Wittgenstein himself wrote in the preface to Philosophical Investigations that the later work "could be seen in the right light only by contrast with and against the background of my older way of thinking."
He needed the early version to make sense of the later one. The revision was not just a correction — it was a development that could only be understood by seeing what it developed from.
This is what version history gives you. Not just what you believe, but the narrative of how you came to believe it. The narrative is not decoration. For ideas that evolve slowly — your theory of management, your model of human motivation, your principles for making decisions under uncertainty — the narrative is the understanding.
Reviewing old versions changes how you think now
Spaced repetition systems like Anki surface material at intervals optimized for retention. But there is a secondary effect that most users discover accidentally: when you review a card you wrote months or years ago, you do not just remember the content. You notice the gap between what you wrote and what you now understand.
That gap is data. It tells you which areas of your thinking have progressed and which have stagnated. It reveals which beliefs you revised based on evidence and which ones shifted through social pressure or mere exposure. It shows you how fast — or how slowly — you actually learn.
Luhmann's Zettelkasten worked this way structurally. Over forty years and roughly 90,000 handwritten notes, he would branch new notes off existing ones, creating internal conversations within the slip box. A note from 1965 might have a response from 1972 and a counter-response from 1980, all linked through his numbering system. The original note was never erased. It became the root of a thread — a visible record of how his thinking on that topic evolved. Luhmann attributed his extraordinary productivity (over 70 books and 400 scholarly articles) not to working harder, but to this system of accumulated, interconnected, versioned thought.
You do not need 90,000 notes to benefit from this. You need the habit of adding rather than replacing. When your belief changes, keep the old version. Annotate why it changed. Let the versions accumulate. Over time, the version history becomes a map of your intellectual development — and that map itself generates insights that no single version can.
Your Third Brain sees what you cannot
When your ideas exist as versioned objects — with timestamps, with changelogs, with the evidence that triggered each revision — AI becomes dramatically more useful as a thinking partner.
Without version history, AI can only work with what you believe now. It has a snapshot. It can summarize it, extend it, challenge it in generic ways. But it cannot tell you that your belief about team management has revised three times in two years without converging, which might signal that you are reacting to surface events rather than updating a deeper model. It cannot tell you that your position on technical architecture has remained unchanged for five years despite three failed projects that should have challenged it.
With version history, the diff between your versions becomes the most interesting input. An AI that can compare your v1, v2, and v3 of a concept can surface the direction of your thinking, identify beliefs that resist revision despite contrary evidence, and ask the one question that matters most for epistemic growth: What would it take to get you to v4?
This is not speculative. The same pattern exists in software: the most useful code review tool is not one that reads your current code — it is one that shows you what changed and asks why. git diff is more useful than git show. The same principle applies to your ideas.
Minimum viable version control
You do not need Git for your ideas. You need a practice. Here is the simplest one that works:
When you revise a belief, do not overwrite the note. Add a new section with a date and the updated position. Keep the original. Between the two, write one sentence about what triggered the change — an experience, a piece of evidence, a conversation, a failure.
Over time, your notes develop stratigraphy — layers of intellectual sediment that you can read like a geological record. The surface layer is what you believe now. The layers beneath are why you believe it, and what you believed before, and what you gave up along the way.
This is what it means to have a knowledge system instead of a collection of current opinions. Opinions are ephemeral. A versioned atom is an idea with a history — and an idea with a history is one you actually understand.
The next lesson — Tags are lightweight relationships — introduces the simplest way to declare that two atoms share something in common. Versioning gives your atoms depth over time. Tags give them breadth across your system.