Research

Latent

A concept expansion engine that explores the hidden structure of ideas — finding analogies across domains, drilling into fundamentals, and bridging disconnected concepts through isomorphic patterns.

Creativity is not random — it's structural

The greatest breakthroughs come from seeing hidden connections between distant domains. Evolution and gradient descent. Version control and memory reconsolidation. Cache invalidation and forgetting. Latent makes these structural analogies discoverable, turning concept exploration from an art into a repeatable process.


How It Works

Seed Concept

Depth

Drill into fundamentals and first principles

Breadth

Find similar patterns across domains

Bridge

Connect two concepts across domains

Deeper insight
Lateral analogy
Novel bridge

Three modes of expansion, each revealing different kinds of hidden structure. Depth peels back layers to find root principles. Breadth finds isomorphic patterns across unrelated fields. Bridge discovers the conceptual pathway connecting two seemingly disconnected ideas. Once the landscape is mapped, other suites can turn it into something operational: Prism can structure and hydrate it, Logic can persist it, and Math can help score or compare what matters most.


Capabilities

Depth Expansion

Drill into the fundamental principles underlying a concept. Peel back layers of abstraction to find the core mechanics and first principles.

Breadth Expansion

Find structurally similar patterns across unrelated domains. Discover that cache invalidation and biological forgetting share the same underlying trade-offs.

Bridge Mode

Given two concepts from different fields, find the conceptual pathway connecting them. Reveal hidden isomorphisms that spark genuine insight.

Multi-Level Exploration

Control expansion depth with configurable levels. One level gives immediate associations; multiple levels reveal deep structural patterns.

Structured Output

Every expansion returns tagged, structured results — not free-form text. Concepts, relationships, and analogies are machine-readable and composable.

Hydration for Downstream Reasoning

Use Latent to build a rich conceptual map first, then hand the result to Prism or Logic so an agent can navigate, hydrate, and reuse the most relevant branches later.

Research Acceleration

Turn hours of literature review and lateral thinking into seconds. Latent surfaces connections that would take a human researcher weeks to find.


See It In Action

// Breadth: find structural analogies across domains
{
  "name": "data-grout@1/latent.expand@1",
  "arguments": {
    "seed": "cache invalidation",
    "mode": "breadth",
    "levels": 2
  }
}

// Bridge: connect two concepts from different domains
{
  "name": "data-grout@1/latent.expand@1",
  "arguments": {
    "seed": "version control",
    "target": "memory reconsolidation",
    "mode": "bridge"
  }
}

Breadth expansion on 'cache invalidation' surfaces analogies in biological memory decay, economic depreciation models, and attention mechanism pruning. Bridge mode reveals that version control and memory reconsolidation share a common pattern: selective preservation under constraint pressure.


Use Cases

Research & Discovery

Accelerate literature review by finding structural analogies across fields. Discover that your problem in distributed systems was solved in evolutionary biology decades ago.

Creative Ideation

Generate novel product concepts by bridging existing ideas across domains. Find the intersection of subscription models and game mechanics that nobody has explored.

Knowledge Synthesis

Build concept maps that reveal hidden relationships in your organization's knowledge base. Understand how seemingly unrelated projects share common patterns.

Latent Space Hydration

Use Latent to expand the idea space, Prism to turn the result into structured summaries or traversable outputs, and Logic to keep the useful concepts available for future agents and sessions.


Composes With

Use Prism to structure, analyze, or render expanded concept maps into agent-ready context. Store durable discoveries in Logic so future sessions can query and traverse them. Bring in Math when you want to rank themes, compare frequencies, or score candidate branches quantitatively. Use Flow to chain multi-step expansion and hydration pipelines.