Linear chat fails at complexity. The Idea Lab replaces it with a 2D canvas where ideas become spatial objects you can see, connect, challenge, and compress.

This guide covers every operational aspect of the Idea Lab. For the underlying cognitive science and architectural decisions, refer to the Spatial Interface Architecture documentation.

---

Getting Started

Opening the Lab

Click IDEA LAB in the top navigation bar. The URL changes to ?view=spatial and the Lab opens as a full-screen overlay. The underlying Brainstorm Reactor chat is preserved in the background — nothing is lost.

To return to the Brainstorm Reactor chat: click the Brainstorm Reactor logo (top left) or press ESC in Zen Mode.

Interface Layout

The Lab is organized into 5 zones:

Creating Your First Card

On an empty canvas, click the Formulate first hypothesis button (or click/double-click anywhere on the workspace). A draft card appears directly below the title.

Two options:

1. Manual input: Type your thought and press Enter. The card is confirmed immediately.
2. AI generation: Write a prompt (e.g., "Analyze the competitive landscape of electric aviation") and press Enter. The AI generates multiple structured cards — each with its own epistemic type, title, description, and confidence score.

You can also click on an existing card and generate from its context. The AI uses the clicked card's content as input context.

---

The Ghost Mechanic

AI-generated cards do not appear as permanent graph members. They arrive as Ghosts — semi-transparent, ephemeral suggestions with a 6-second countdown.

The Ghost mechanic is a deliberate intervention against Algorithmic Cognitive Fixation — the empirically documented tendency to accept AI suggestions uncritically when they are presented as finished artifacts.

Cognitive Forcing: Edge-Type Picker

When confirming a Ghost that has a parent relationship, the system asks you to explicitly classify the type of relationship:

• ✓ Supports (green)
• → Generates (blue)
• ⚡ Contradicts (red)
• ◎ Specifies (cyan)

This forces conscious evaluation of logical relationships rather than passive acceptance.

---

Cards and Connections

5 Color Families

All 14 card types are grouped into 5 intuitive color families for instant visual parsing:

Shape Language

Card shapes encode epistemic character without relying on color alone:

• Structured (8px radius) — Stable, verified types (Fact, Constraint, Assumption)
• Exploratory (12px radius) — Open, flexible types (Hypothesis, Question, Analogy)
• Organic (asymmetric: 20px 6px 20px 6px) — Disruptive, creative types (Wildcard, Provocation, Disruptor)

Card Actions

Hover over any confirmed card to reveal the action menu:

Inference Audits & S-Nodes (ASPIC+ / AIF Standard)

Clicking the Inference Audit button (GitCommit icon) on any card executes a formal logical audit. The engine analyzes implicit premises and unfortified inference steps, generating dedicated Inference Nodes (S-Nodes):

• inference_rule (Indigo): Explicitly highlights a logical inference step and rule application.
• inference_conflict (Purple): Exposes an underlying logical conflict or undercut in reasoning.

Context-Sensitive Magnetic Drop-Zones

When dragging cards across the canvas, target cards within a 150px magnetic proximity radius dynamically emit an indigo glowing aura (aura-active). Releasing the card inside the active drop-zone automatically binds the nodes using typed inference edges (infers or undercuts).

Macro-Skeleton Zooming

The canvas automatically adapts node presentation across 3 zoom thresholds:

• Micro Zoom (> 75%): Full content display with rich text and scrollable descriptions.
• Meso Zoom (40%–75%): Compact card view with 3-line truncated copy.
• Macro Zoom (< 40%): Auto-suppresses text copy to render glare-free topological skeleton nodes, enabling rapid structural auditing of large semantic graphs (50+ cards).

Creating Connections

Drag from the connection handle (small circle at the card edge) to another card. The connection defaults to generates — you can change the type via the context menu.

Multi-Selection

Shift+Click or Cmd/Ctrl+Click on multiple cards. Selected cards receive a visible ring indicator. Actions available:

• Drag all selected cards simultaneously
• Merge (Genetic Crossover) — fuses the selected concepts via AI
• Group — wraps them into a collapsible Macro Node

---

View Modes

The floating bar at the bottom of the canvas provides 7 layout algorithms. Click any icon to instantly rearrange the graph.

Most layouts require at least 2 confirmed cards.

Method Analyzer (DNA Icon)

Beyond the 7 standard layouts, the Method Analyzer popover (DNA icon in the view bar) applies 5 structured analytical frameworks:

All transformations are non-destructive and instantly reversible. Ghost Trails (dashed SVG lines) show where each card was moved from, fading after 1.2 seconds.

---

Semantic Zooming

Cards display different detail levels based on zoom:

Density can also be toggled manually via the Controls panel (bottom left).

---

Filters and Lenses

Meta-Filter Panel (Top Left)

Contextual Lenses

Unlike filters (which hide cards), lenses dim non-matching cards to 15% opacity while keeping everything visible.

Keyboard: Hold the key to activate, release to deactivate. This creates a natural "X-ray" effect — you glance through an analytical filter and return instantly to the full picture.

Click: Click the lens button in the filter panel to toggle it on/off persistently.

---

Analytical Tools

The vertical tool rail on the right edge provides 4 instruments:

Stress Test (Adversarial Audit)

Icon: Shield | Requires: 2+ confirmed cards | Cost: 1 AI call

The AI evaluates every single edge and node in your graph:

• Edges receive an integrity score (0.0–1.0). Edges below 0.5 visually "erode" (dashed line, reduced opacity).
• Nodes receive a fitness score (0.0–1.0). Nodes below 0.3 are dimmed with a warning badge. Nodes above 0.8 receive a gold shimmer (cornerstones).
• BFS Propagation: If a foundational node scores poorly, all downstream children inherit degraded fitness — the erosion cascades visually through the entire graph.

Blind Spot Scanner (Eigen-Shadows)

Icon: Eye | Requires: 2+ confirmed cards | Cost: 1 AI call

Uses null-space projection (Modified Gram-Schmidt) on your cards' 384-dimensional embeddings to find the concept maximally distant from everything currently on your canvas. Spawns a disruptor ghost card representing a literal cognitive blind spot.

Genetic Crossover (Merge)

Appears: Only when 2+ cards are selected | Cost: 1 AI call

Fuses elements from all selected parent concepts into one highly divergent synthesis node. Automatically linked to parents via mutates_from edges. This is bisociation — the generation of emergent properties, not summarization.

Grouping (Macro Nodes)

Appears: Only when 2+ cards are selected | Cost: None

Wraps selected cards into a collapsible group. Groups reduce visual density without losing information. They can be expanded or dissolved at any time.

---

Time Travel

Opening the Timeline

Click the clock icon (🕐) in the view mode bar. The Time-Travel slider appears.

How It Works

The Lab maintains a ring buffer of up to 50 structural snapshots. Each snapshot captures all cards, positions, connections, and metadata. Snapshots are automatically recorded when:

• Cards are added or archived
• Connections change
• A method transformation is applied

Scrubbing

Drag the slider left to scrub into the past. The entire canvas desaturates proportionally:

The desaturation is a deliberate signal: the past is immutable.

Restoring

Click Restore (amber button) to push the scrubbed state as a new snapshot. The timeline closes automatically. This is non-destructive — the previous state remains in the buffer.

Constraints

• No scrubbing during active AI generation
• Snapshots are client-side only (not cloud-persisted)
• Ring buffer overwrites oldest snapshots after 50

---

Export and Sharing

Click the share icon (📤) in the Navbar. The export menu is divided into 3 sections:

Section 1: → Brainstorm Reactor Pool

Smart Export Templates:

The template suggestion engine runs entirely client-side (zero API calls).

Section 2: Share

Share Link creates a public read-only link. If the lab has genesis snapshots (auto-captured during creation), the recipient sees a 5-second time-lapse of the canvas building itself before the final state loads.

Section 3: Files

Section 4: Focus Text Editor (In-Lab Synthesis Drawer)

Click the document icon (FileText) in the left workbench rail to slide open the integrated Focus Editor drawer directly alongside your 2D canvas:

• Distraction-Free Synthesis: Write notes, distill findings, or structure your project specification without leaving the spatial environment.
• Instant Card Ingestion: Select one or more cards on the canvas and click Karten einfügen (Insert Cards). The titles, epistemic types, and descriptions are cleanly appended at your current cursor position as formatted markdown.
• Local Persistence: The draft autosaves locally in your browser (reactor_lab_focus_editor_draft), persisting across canvas navigation and reloads.
• Round-Trip to Reactor: Click In Reactor Chat senden to transfer your synthesized draft straight into the main Reactor input prompt for subsequent agentic workflows.

---

The 50-Card Barrier

A hard density limit enforces cognitive discipline. At 50 confirmed cards, further AI generation is blocked and the system forces a Smelting sequence.

Smelting Process

1. SmeltingModal appears (400ms entrance delay)
2. Select 3–20 related cards (Shift+Click)
3. Click "Merge" → AI synthesizes a single Interface Node (type: synthesis)
4. Original cards are archived, Interface Node takes their place
5. Canvas drops below limit — generation resumes

Interface Nodes represent compressed sub-graphs: an entire cluster's logic encoded in a single dense card.

Prevention Strategies

• Archive weak cards early (action menu → Archive)
• Run the Stress Test — cards with fitness < 0.3 are archive candidates
• Use Smart Export intermittently to send intermediate syntheses to the Pool
• Group related cards into Macro Nodes to reduce visual density

---

Acoustic Cognition

Activation: Click the audio icon (🔊) in the Navbar.

The Lab provides an optional ambient sound layer that dynamically maps to your graph state:

Zoom controls reverb: Zooming out increases reverb time (large room feel), zooming in decreases it (intimate, dry sound). This is "Ecological Zooming" — you hear the spatial scale of your current focus.

Safety: Volume capped at 25%. Lowpass filter at 1400 Hz prevents high-frequency listener fatigue. Use headphones for the best experience.

---

Smart Friction: Autonomous Operators

12 operators analyze your canvas and intervene when cognitive biases are detected. All run client-side — zero AI calls, pure graph analysis.

---

Zen Mode

Press ⌘. (Cmd+Period) to hide all UI elements — Navbar, toolbars, filter panel. The canvas fills the entire screen. Press ESC to return. On mobile, a floating exit button ensures touch accessibility.

---

CAD Score

The Canvas-Aware Diversity score in the Navbar measures epistemic variety:

---

Quick Reference: Keyboard Shortcuts

---

Quick Reference: All 13 Edge Types