Knowledge

Knowledge is your workspace’s knowledge base: a connected store of audience definitions, offers, messaging, and the insights every campaign plays back into it. Knowledge is the single home for everything your workspace has learned about your market and about itself — three object types live inside it: Assets, Learnings, and Memory (Speicher).

Assets

An Asset is a typed, versioned building block: ICP, Persona, Offer, Value Prop, Signal Set, Messaging Angle, Positioning, Proof. Assets are referenced, not copied by Playbooks and columns — when an asset changes, the new version takes effect wherever it was approved and pinned, not everywhere at once without asking.

Learnings

A Learning is an evidenced insight from live campaigns — a winning copy pattern, an objection handle, a conversion profile, a channel or signal effect. A Learning is structured (typed fields: pattern, outcome, confidence, why it worked, evidence count, source conversations) and referenceable like an Asset — not a block of free text. It carries a confidence score that strengthens with further evidence or fades over time (reinforcement/decay): a learning backed by a single data point counts less than one backed by twenty.

Learnings are their own kind of Asset — they inherit the same machinery: versions, approval, pinning to a playbook, referenceability in prompts and analyses. The system distills them from what actually books meetings and generates replies, and drops them as a proposal into an approval queue inside Knowledge. Only after your approval does a Learning flow into the copy and qualification prompts. Details in Learning Loops.

Memory (Speicher)

Memory — Speicher — is your workspace’s self-maintaining memory in text form: “how we build campaigns,” “user preferences,” “mistakes & corrections,” “debugging agents & chains.” Where Learnings are the revenue-relevant, structured insights, Memory holds the procedural knowledge — how work gets done in this workspace, which tool for what, what already went wrong once. Agents read Memory into their context, write updates back after a session, and a hygiene routine merges duplicates and fades what’s stale — the memory maintains itself over time.

Two learning loops, one home

Knowledge carries two optimization loops that deliberately don’t mix:

1. Learning you and the craft — so the system doesn’t repeat the same mistakes and gets better over time at building end-to-end campaigns, table references, and agent chains. This loop lives in Memory (markdown, self-maintaining).
2. Learning revenue — which offer, which ICP, which signal, which persona, which message, which channel books meetings and generates replies, so the winning pattern gets replicated. This loop lives in Learnings (structured, approval-gated) and feeds prompts, agent chains, and analyses from the same typed records.

Rolling out

This consolidated view is rolling out in stages. Today, Knowledge already lives in your workspace — as persona and ICP definitions per playbook, as confidence-scored learnings from the running learning loop. Coming next: the structured Learning assets with an approval queue (first stage already built, togglable per workspace), Memory as browsable, self-maintaining documents, and the fully connected, workspace-wide browsable Knowledge view.

The Sales Blueprint doesn’t become a third, competing store — and it isn’t being retired. It stays the early understanding funnel: it ingests sources (uploaded documents, CRM analytics, outreach performance, and connected knowledge integrations like Notion or a generic knowledge MCP) and proposes matching Knowledge objects from them (ICP, Persona, Offer, Messaging → Assets; winning patterns → Learnings), which land in the same approval queue inside Knowledge. Knowledge stays the single structured runtime source that prompts, agent chains, and the closed loop read from — the Blueprint is the intake funnel that feeds it, not a competitor to it.

See also: Playbooks bind pinned asset versions; Learning Loops describes how Learnings and Memory entries arise.