The three memories

A grounded worker draws on three different kinds of memory, and askTheodor keeps them deliberately separate so your trusted facts never get muddled with a worker’s scratch notes.

What you’ll learn

• The three memory layers a worker can read from.
• Why keeping them apart makes your workforce easier to reason about.
• That all three live on your machine and are searchable.

The three layers

1. Private memory — what a single worker quietly learns and recalls during its own runs. It’s personal to that worker: lessons, preferences, and facts it picked up on the job. Nobody else reads it.
2. The tribe memory — shared within one Company. It’s the team’s common ground, so workers in the same neighborhood pull from the same context rather than each starting cold.
3. The Library — your deliberately curated knowledge base of documents, notes, lists, links, and files. You decide what’s on the shelf; every relevant worker can reach for it.

Why separate them?

The Library is your reference material — trusted, on purpose. Private memory is a worker’s own notebook. Mixing them would let one worker’s guess contaminate your facts. Separation keeps each layer honest.

All local, all searchable

Every layer lives on your machine — askTheodor is local-first. And each is searchable, so a worker can find what it needs the moment a task calls for it.

What the separation buys you

The three layers aren’t bookkeeping — they’re what stops an AI team getting confused about what’s true. Three moments where it shows:

Onboarding a second worker takes minutes, not a re-brief. Put your pricing, positioning and tone in the Company’s shared layer once. The next worker you hire into that Company starts with the same common ground rather than an empty head — you skip the “here’s how we do things” conversation entirely.

A worker’s wrong guess stays contained. If a worker infers something slightly off during a run, that inference lives in its private notes. It doesn’t get promoted into your Library and it doesn’t quietly become the team’s version of the truth. When you correct the Library, you’re correcting the authoritative copy — because there’s only one.

You can answer “why did it say that?” Because the layers are distinct, an odd answer has a findable source: your curated document, the Company’s shared context, or that worker’s own notes. That’s the difference between debugging a workforce and shrugging at it.

Recap

Workers think with three memories: their own private notes, their Company’s tribe memory, and the shared Library you curate. Keeping them distinct is what makes a worker both grounded and trustworthy. Next, we open the Library and see how a worker actually pulls from it.

📖 Reference in the docs: Worker profile

➡️ Next: The Library & RAG