You read a piece three months ago that introduced a way of thinking about a problem you keep running into — something about second-order consequences, or maybe expected value, or maybe the local-versus-global optimum. The piece was good. You highlighted the relevant section. You probably saved it. And right now, when the same problem is sitting on your desk and you'd give a lot to have that framework in front of you, you cannot find it.
Most people who collect mental models and thinking frameworks run into this version of the problem. The collection grows — books, podcasts, essays, courses — but the act of retrieving the right model when the right moment shows up never quite gets built. You end up with a stack of highlights you'll re-read someday and a set of decisions you make from gut.
A vault that holds the models you've actually internalized (or want to), with examples from your own life and an agent that can surface the right one when you ask, fixes most of it.
What's worth tracking and what isn't
The first mistake is collecting every framework you encounter. There are hundreds — first principles, inversion, opportunity cost, Hanlon's razor, Chesterton's fence, the OODA loop, expected value, base rates, the tyranny of the marginal user, Goodhart's law, Cunningham's law, the Peter principle, circle of competence, on and on. Most of them are useful in narrow situations. Trying to maintain notes on all of them is a research project, not a thinking practice.
The version that works is tracking the ones you actually find yourself reaching for, with examples from your own life that show when and how you used them. A library of fifteen models you genuinely internalize beats a database of two hundred models you've highlighted but never applied.
The signal that a model belongs in your vault is that you've used it at least once and want to be able to use it again. Not "this seems clever." Not "I should think this way more." Just: "I needed this last week and I'd like to have it findable next time."
One page per model, with examples from your own life
In Docapybara, each model gets its own page under a Models & frameworks parent. Pages nest with no depth limit, so you can group by domain if it helps — Decision-making, Problem-solving, Negotiation, Risk — or just keep one flat list and let the agent search across.
The page structure is simple:
- The model in plain English. Two or three sentences in your own words. If you can't write it in your own words, you don't actually understand it yet.
- When it applies. The specific situations where reaching for this model is appropriate. As specific as possible.
- When it doesn't. The situations where it'd lead you wrong, or where it's not actually the right framing.
- Examples from your own life. Two or three real moments where you used the model (or wish you had). Dated. Brief.
- Source. Where you first encountered it — book, essay, podcast, conversation. Link or title.
The "examples from your own life" section is the one that turns a model from a piece of trivia into a tool. Generic examples ("imagine you're choosing between two job offers…") don't help future-you. Your examples ("the [specific decision] in March 2024 — used inversion, asked what would have to be true for this to fail") become the reference material.
For the broader habit of building a personal knowledge structure that compounds, see How to Build a Personal Knowledge Wiki Without Trying. The mental models library is one of the highest-leverage threads inside the wiki.
Capturing models from books and essays you read
Most models enter your library through reading. The piece-of-friction is that highlighting in a Kindle or a browser doesn't translate into anything you'll actually reach for later. The fix is a small habit: when you encounter a framework that seems applicable to your life, give it a short page within a week of reading.
The page doesn't have to be long. A paragraph in your own words plus one example from your own life, plus the source. If you can't write the paragraph or recall an example, the model didn't actually land — let it go and revisit if it shows up again.
For book-length material, a separate Book summaries habit helps the models from each book stay associated with the book they came from. See Book Summaries and Reading Lists: A System That Actually Lets You Recall a Book for the broader shape; mental models extracted from books become individual pages in the Models library, with a link back to the book summary.
For models from podcasts or talks, voice notes immediately after capture the framework while it's fresh. "From the [podcast] episode this morning — the framing about second-order effects on team incentives. Specifically the example about [scenario] resonated. Want to draft a page on this." The transcript becomes the seed for the eventual page.
Asking the agent to surface the right model
The reason this library pays off is that the agent can read across all of it when a real situation comes up. You're sitting with a problem and you ask: "I'm trying to decide whether to accept the consulting engagement next month — what models in my library might be relevant?" The agent reads across your pages and surfaces the candidates with brief explanations of why each might apply.
This is genuinely different from searching a generic mental-models website. The agent surfaces your models — the ones you've actually internalized, with the examples you've attached — which means the application to your specific situation is grounded in your own experience.
Useful prompts:
- "For the situation [describe], pull every relevant model from my library."
- "What models have I been using most often this year, based on the example pages I've added?"
- "Find models that would have been useful for the decisions I journaled about in the past three months but didn't use."
The third one is the most powerful and most uncomfortable. It pairs with How to Build a Decision Journal in Your Notes App — your decision journal records what you actually thought at the time; the models library can show what frameworks you had available but didn't apply. Over time, that's a real form of learning.
When a model gets challenged or refined
The other discipline that keeps the library alive is updating models when your understanding shifts. Most frameworks are simplifications. Most simplifications are useful in some situations and misleading in others. The version of a model you write down at year one is rarely the version you'd write at year five.
When a real experience challenges a model — you applied expected value and the right answer turned out to be loss aversion, you applied first principles when Chesterton's fence would have served better — note it on the model's page. "April 2026 — used this model on [decision] and it pointed me wrong because [reason]. Adding a 'when it doesn't apply' note to refine."
This is the part that separates a library from a museum. The pages get edited. The "when it doesn't apply" sections grow. The examples from your life accumulate. The library matches your actual thinking instead of the static version of someone else's.
Frameworks for specific domains
Some frameworks are general-purpose; some are domain-specific. Expected value applies broadly; the AARRR funnel applies to product growth; the OODA loop applies to fast-moving competitive situations; Liskov substitution applies to object-oriented design. The general-purpose ones live in the main library; the domain-specific ones can live in their own subsections.
For people whose work touches multiple domains — generalist consultants, fractional executives, founders wearing many hats — this organisation matters more. The library becomes a tool for matching the right framework to the right context. "I'm advising a client on a pricing question this afternoon — what frameworks in my library are relevant for pricing decisions in B2B SaaS specifically?" The answer comes back narrowed.
For the broader version of "having the right context for many different engagements," AI Notes for Fractional Executives Managing Multiple Engagements covers the engagement-management shape; the models library sits inside it as the cross-engagement thinking toolkit.
A starter shape that grows from real use
If you're starting a models library this week:
- A Models & frameworks parent page.
- A page for each of the three or four models you reach for most often. Plain-English explanation, when it applies, when it doesn't, examples from your life, source.
- A small habit: when a model in a book or essay actually lands, give it a page within a week. Otherwise let it go.
- A monthly review: ask the agent which models you've been using and which need refining.
That's it. No taxonomy, no template, no completeness goal. The library grows from real use; it stops growing when you stop reaching for new models, which is fine.
The point isn't to be a person with two hundred frameworks. It's that the small library you genuinely internalize becomes part of how you think — and the agent makes the right one findable when the right moment shows up.
Try Docapybara free — start with the model you wish you'd had in front of you last week, write the page in your own words, and the rest of the library will grow as your thinking does.