Knowledge base
A knowledge base is a body of material — documents, notes, references — kept together so that questions can be answered from it later, rather than from memory or from a fresh search.
- Also called
- Second brain, PKM system
- Usual failure
- Filing outpaces the payoff
- Honest test
- How much can you ask about?
The word carries an assumption that only holds at work
In its original setting the term is organisational: a support team’s answer library, a company wiki. Someone’s job is to maintain it, and that job is why it stays usable.
Carried over to one person, the same word quietly keeps the maintenance assumption and drops the maintainer. So the personal version is sold as a system to build — tag things, link things, write summaries, keep an index — and the cost structure of that is brutal in a specific way: the work is due immediately, and the payoff arrives only if you later ask the question you were filing against. Most filing is never repaid.
The work is due on day one; the payoff only arrives if you later ask the question you filed for.
The same year, two ways. The dashed cells are where build-first systems usually stall.
The result is uniform enough to be predictable. Month one is well-organised. Month three has a folder called “to sort” with more in it than the organised part. Nothing is broken; the system simply stopped being worth its upkeep, and the material is still there, unfindable.
Which suggests the honest test for any of these tools. Not “how good is the structure” but: what fraction of what you have can you get an answer out of, without remembering where you put it? If the answer depends on your own filing discipline, you have bought a filing cabinet and called it a knowledge base.
Worth saying plainly: a searchable pile is not understanding. Being able to retrieve a passage is a lower bar than having read the thing, and no amount of retrieval quality closes that gap.
How MindMax handles it
There is no building step. MindMax opens a folder you already have — an Obsidian vault, a downloads folder of PDFs, three years of notes — and works on it in place. Nothing is imported, nothing is converted, and nothing has to be tagged before it is useful.
Everything visible in that folder is searchable by default. You do not mark documents as included, and there is no separate index to keep current: material you drop in is material the AI can reach. What you want kept out, you hide from the AI — and it still shows up in your own searches. When you want particular documents read word for word rather than searched, you add those to the conversation; the rest of the folder stays reachable behind them, and the answer says which part came from where.
Inclusion is the default; exclusion is one right-click. Hide from AI keeps a file out of the AI’s reach without moving it.
Common questions
Why do personal knowledge bases fail?
Because they are sold as systems to build. Filing, tagging and linking cost effort the day you do them, while the payoff only arrives if you later ask the question you filed for — so most of the effort is never repaid, and the backlog of unfiled material grows until the system stops being used.
Can I make an AI knowledge base from my own files without organising them?
Yes. MindMax opens a folder you already have — PDFs, ebooks, Word files, notes — and everything visible in it is searchable by the AI without tagging or importing. You can hide individual files from the AI if you want them left out.
Is a searchable knowledge base the same as understanding what is in it?
No. Retrieving a passage is a lower bar than having read the document. Search helps you find material; reading, annotating and questioning it is still the work.