Skip to main content
As you work through research, you build up knowledge — conclusions from experiments, insights from analysis, facts you’ve verified. Qualia’s Knowledge system captures this knowledge as claims with clear provenance back to evidence.

​
What are claims?

A claim is a durable fact or conclusion tied to your project:
  • Content: The actual statement or finding
  • Sources: Evidence supporting the claim (code cells, files, other claims)
  • Type: How the claim was created
  • Recall priority: How readily agents should see it again
Claims are project-scoped — they belong to a specific project and persist across sessions. The one exception is global recall priority, described below.

​
Types of claims

Evidence claims come from three capture paths — notebook capture, script capture, and file capture. The Knowledge sidebar presents all three as one Evidence type, with the capture source (Notebook, Script, or File) shown alongside each claim.

​
Runtime-captured

Created when code runs and produces a value, read straight out of the running process:
  • In a notebook, variables marked as “required” on a task are captured automatically, and the claim links to the cell that produced them
  • In a script, the qualia client submits the value as the script runs; the claim links to the line of the script that produced it
  • Either way the value is read from the live runtime, so it keeps its real precision and type — it is never something the agent typed in
  • These claims mirror the evidence — when the cell or script changes, the claim is revalidated

​
File-captured

Created when a durable result lives in a text file that the agent’s work produced — a results file, a log, a CSV or JSON output, or a specific line of a script:
  • The agent quotes the exact span; the backend verifies the quote appears verbatim in the file and records the path and quoted text
  • The claim is just that quote — no agent-written summary or commentary. The file path and line range are shown alongside it
  • The quote has to identify one place in the file: if the same text appears more than once, the capture is rejected until the agent quotes enough context to be unambiguous
  • A single claim can cite several files only when they are two views of the same fact — for example a result and the code that produced it
  • Every time the claim is read, the quote is re-checked against the file. An edit that moves it updates the line range, an edit that removes it marks the source as changed, and a file that was moved or deleted marks it as unreadable
The value is anchored to the file, not typed in freely: a quote that isn’t present in the file is rejected. You can create file-captured claims yourself, too — see Saving a selection as evidence.

​
Synthesized

Written by agents to summarize understanding:
  • Combine multiple sources (cells, files, other claims)
  • Represent conclusions drawn from evidence
When an agent works on a task, it often creates a synthesized claim before marking the task complete. This provides a record of what we learned by working on this task.
Sources are directional: they say one claim rests on another. Some claims simply belong together without either one supporting the other — two findings that examine the same subject from different angles, a result and the limitation that qualifies it, a figure and the conclusion it illustrates. Agents can record that kind of link as a related connection:
  • The link is nondirected. Neither claim is presented as evidence for the other, and it never affects validation or the provenance chain.
  • Both claims list each other, so opening either one shows the connection.
  • Agents see a claim’s related items whenever they look that claim up, which is how a later agent discovers context it would otherwise miss.
Related claims appear in a Related section of the claim detail panel, and as a dashed line — with no arrowhead, unlike a source edge — in the graph view.

​
Bringing existing work into knowledge

Work that predates Qualia — old notebooks, scripts, and results files — can be turned into claims in one pass. On the new chat screen, choose Convert existing work into Qualia knowledge and pick a file or a folder. An agent reads what’s there, captures the concrete results it finds as file-captured claims, and writes findings on top of them. The path has to sit inside your workspace, because every claim points back at a workspace path that gets re-verified whenever the claim is read. Binary files are skipped: a PDF or an image has no text to quote.

​
Recall priority

Every claim carries a recall priority — the agent’s judgement of how readily future agents should see it again. It describes the claim’s standing, not how confident anyone is that it’s true, and not how relevant it is to whatever an agent happens to be doing right now. It applies to all knowledge, so a captured value can outrank a written finding. Most knowledge is Low, and that’s the intended default — it means “recall this when it matters”, not “this doesn’t matter”. Reserve Global for facts that hold across the whole workspace; a result that’s specific to one dataset or experiment belongs at High. Global and High claims are injected into the AI’s context automatically, so a finding an agent files or revises reaches the other agents at their next step without anyone re-explaining it. Every claim shows its priority as a badge in the Knowledge sidebar, and you can sort the list by priority.

​
The Knowledge sidebar

Open the Knowledge sidebar from the activity bar to browse your project’s claims.
  • Type filter: Show only evidence, findings, tables, or figures. Evidence can be narrowed further by its capture source — Notebook, Script, or File — via the indented checkboxes under it
  • Search: Find claims by content or ID
  • Sort: Newest/oldest, or by recall priority

​
Viewing claims

Each claim card shows:
  • Claim content (truncated for long claims)
  • Type indicator
  • Recall priority badge
  • Source count
Click a claim to see full details and navigate to sources.

​
Views

  • List view: Claims as cards
  • Graph view: Visual DAG of claim relationships

​
Working with claims

​
Saving a selection as evidence

Agents aren’t the only ones who can file evidence. Select any text in a file, a notebook cell, or a text output, and a Save as evidence button appears next to the end of the selection once you release it. It’s also available as Save Selection as Evidence in the menus and the command palette. Saving opens a short form:
  • The selected text is shown but not editable — it’s the record, and it’s stored exactly as you selected it
  • A caption is optional. Use it for what the text means; it becomes the headline for the claim, with the quote shown beneath it. The quote itself stays untouched, so an agent reading the claim can always tell your words from the source’s
  • A recall priority controls how readily future agents see it, using the same scale as everything else in the graph. Each level explains itself in the picker; Global additionally surfaces the claim to every agent in the workspace, always
The selection is verified the same way an agent’s quote is: the backend reads the file or cell back and refuses the capture if the text isn’t there verbatim, or if it appears more than once and so doesn’t point at one place. If the source changes after you save, the claim reports it as changed rather than quietly showing a stale citation. Evidence you capture this way is marked Captured by you on its claim card, so it stays distinguishable from evidence an agent filed.

​
Validation

Claims can be validated against their sources:
  • Validate: Check if the claim still holds given current evidence
  • Validate upstream: Check this claim and all claims it depends on
Validation runs a check — if sources have changed significantly, the claim may be automatically updated.

​
Provenance

Every claim links to its sources:
  • Notebook cells: Jump to the cell that produced evidence
  • Files: Open the referenced file
  • Other claims: Navigate the claim graph
A source that quotes a specific span also shows which half of a cell it came from and its line range, and flags itself when the text it quoted is no longer there. This chain of evidence lets you trace any conclusion back to raw data and code. Related claims are deliberately kept out of it — they are associations, not evidence, so they never appear as sources.

​
Claims and tasks

The knowledge system connects to tasks:
  • Tasks require synthesized claims for completion
  • Claims link back to the task that produced them
  • The task graph and claim graph together document your research path

​
Task-variable integration

When tasks pass variables to downstream tasks via task-variable evidence (the qualia package), those values become runtime-captured claims. This creates a documented chain of data flow through your project.