Vinc

What to add

Vinc will take anything, but putting in just anything helps with nothing. What actually pays off is material you keep coming back to. The three below are the common cases that meet that bar.

When it is worth doing

There is one test. Have you already read it, and are you looking for it again?

Material you read once and never return to has little reason to be here. A search engine already does that well.
What this tool is for is the opposite: material you cannot remember the location of every time you need it, and material scattered across documents that you have to re-thread by hand.
Most of the value is in the connections. Ten documents are fine in a folder; two hundred that reference each other need a map.

If you are not sure, import an official pack from the community gallery and look at the shape it makes. Five minutes is usually enough to tell.

1. Your development work

Code tells you what it does, but not why it ended up that way. Gather the places that reasoning lives (design notes, PR descriptions, incident write-ups, specs) and the version of you six months from now does not relitigate the same argument.

Put in: design decisions and their reasoning, incident retrospectives, architecture notes, third-party API contracts, anything that records "we did it this way because".
Leave out: the code itself. It lives in the repo and changes constantly. What belongs here is what the code does not explain.
Where it pays off: you ask an assistant "why is this module shaped like this?" and get back the decision your team actually made rather than a generic answer.

With a connector you can ask from inside the editor. Cursor, Claude Code, and Codex connect with one command.

2. Your own records

Books you read, talks you watched, notes you wrote, meetings, trips. The problem with personal material is not volume, it is not finding it again. You cannot remember where you wrote it down, so you end up searching from scratch.

Put in: what you wrote yourself, first. Reading notes, meeting minutes, journals, lecture summaries. Your own sentences match far better than someone else's prose.
Why it works: meaning search finds things without matching words. You can pull up "that thing I read about negotiation" from a half-memory.
Privacy: the desktop app runs entirely on your machine. Create no account and nothing leaves it. For personal records that is usually the deciding factor.

Sync only uploads the topics you choose. You can keep a journal local and put only work notes in your account.

3. A subject you are learning

The hard part of a new field is not memorizing terms, it is grasping how they hang together. Build it up as concepts for nodes and relationships for lines, and that structure becomes something you can look at.

Put in: one document per concept. Write the relationships to other concepts rather than the definition: "A is a prerequisite for B", "C and D solve the same problem differently".
What the map tells you: the 2D view lays things out in dependency order, so you can see what has to come first. A point with almost no connections is a concept that has not found its place yet.
It travels: a well-organized set of concepts can go out as a pack, and you can pull someone else's into your own map.

The assistant does the structuring. Point it at your material and ask it to organize the concepts and relationships into Vinc. Vinc stores what it is given and never guesses at meaning. If you want it filed a particular way, write that down once in your conventions.

Mixing them is fine

The three are split for explanation, not as walls. In practice one map holds work decisions and personal notes together, separated by domain. They get different colors in View, and search narrows with domain:.

Often the connections that cross fields are the interesting ones. Points that join different colors are called gateways, which is also why gallery cards show a cross-domain share.

Where to begin

Take the 20 to 50 documents you most recently thought "where did I write that down" about, and import them.
Look at View for where it clusters and where it is empty. Empty is what you have not put in yet.
Connect an assistant and ask it the questions you normally ask. When the answers start being different, add more.