ThinkingRoot Docs
Core Concepts

How ThinkingMemory works

Plain language for sources, claims, citations — the model that makes everything else click.

Most “knowledge” tools dump text into a search index and hope the model finds the right chunk. ThinkingRoot stores claims: short facts, each locked to the exact place you taught them. That graph is ThinkingMemory — what your AI reads when it answers.

That’s why answers can show citations — and why “I don’t know” is allowed.

How ThinkingMemory works
1

Teach

Upload docs, notes, or a URL.

2

Remember

ThinkingMemory locks facts to their source.

3

Ask

Get answers with citations — or an honest “I don’t know.”

Nothing is invented. If it isn’t in what you taught, the answer stays empty.

The chain of custody

Teach  →  Compile  →  Claims + sources  →  ThinkingMemory
                                              │
                         Ask / recall ◀───────┘
                         (answers cite the claims)
  1. A source is what you upload (file, note, URL) — tracked so changes are visible.
  2. Compilation turns it into claims (facts) with provenance (where each fact came from).
  3. Claims mention entities (people, products, ideas) and how they connect.
  4. When you ask, ThinkingMemory finds the right claims and builds the answer from them — or stays empty.

Terms, in one line each

TermWhat it is
SourceSomething you taught — a doc, note, or URL.
CompilationTurning sources into claims ThinkingMemory can use.
ClaimOne atomic fact, locked to its source.
WitnessThe byte-level proof behind a claim.
EntityA named thing — person, system, concept.
RelationHow two entities connect, backed by claims.
BranchA safe fork of ThinkingMemory — explore, then merge.
HealthChecks for contradictions, stale facts, grounding gaps.

Why this matters

An answer isn’t “the model’s opinion.” It’s a set of citations you can open. And because ThinkingMemory knows what it contains, “I don’t have that” is a real, first-class answer.

Read on for each piece in depth.