Building a second brain.

The "second brain" framing suggests a large, structured system for capturing everything you know. Almost nobody sustains one. What most people can sustain is a much smaller pattern: one Project per active area of life, with three files inside each, updated weekly.

What goes in each Project.

Three files, no more. Adding more is where these systems fail.

**Voice** — a short document (500-1500 words) capturing how you think about this area. Not a summary of what you know; a summary of your approach. For work, this might be your management philosophy. For a research area, this might be your theoretical commitments. For a relationship, this might be your patterns and what you're working on.

**Facts** — a running document of the specific things you want to remember. Names, dates, decisions, references. Structured however works — bullet list, chronological entries, tables. This is the "brain dump" file.

**Context** — the pointers to relevant external material: URLs, book titles, papers, docs. Not the content — just the pointers, so LADLE can pull them if needed but you don't fatten the Project unnecessarily.

Why exactly three.

Because two is too few (you conflate voice with facts) and four is where entropy starts to win. Three files fit in your head as an organizational structure and each one has a clear role that doesn't overlap.

The weekly update.

Once a week, spend 10 minutes updating one file in each active Project. This is the whole discipline. If you're not doing this update, the second-brain framing is aspirational and you shouldn't pretend otherwise.

How LADLE uses the Project.

Every chat you start inside a Project sees all three files as context. When you ask a question, the model draws on your voice document for how you approach the area, on your facts document for the specific details, and can pull from context URLs when relevant.

The compound effect: over three months, chats inside a Project produce dramatically better responses than fresh chats, because the model isn't reasoning from generic priors — it's reasoning from your specific pattern of thinking about this area.

What not to do.

Don't try to run one Project for "everything". The point of Projects is scope; a Project called "life" has no scope and its voice document devolves into meaningless generality.

Don't include chats in the Project's shared context. Chats are for interaction; files are for accumulated knowledge. Mixing them makes both worse.

Don't backfill years of notes into a new Project. Start empty. Add material as it becomes relevant. The value comes from what you use, not what you've stored.

The sustainable end state.

For most people, three to six active Projects at any time. Some persist for years (your professional voice, your ongoing health context). Some run for a few months (a specific project at work, planning a trip). Retire them when they stop being active — deleted Projects are irretrievable, so export first if the knowledge might matter later.