← All agents

Live · Knowledge & Insight Agent

Cornelius

AI second brain that captures insights, discovers cross-domain connections, and compounds knowledge.

01

Role

Knowledge & Insight Agent

02

Status

Live

03

Time to live

4-6 weeks

Where it breaks today

You have original insights during conversations, meetings, and research — then they vanish. Your notes sit disconnected. Patterns across domains remain invisible. The knowledge that makes you irreplaceable lives scattered across tools, conversations, and memory. Your best ideas disappear. Your knowledge stays siloed.

The fix

Cornelius actively hunts for your unique thinking, captures it in a persistent knowledge graph, and discovers connections across domains that semantic search alone would miss.

What we build

An agent is a stack.

01

Conversational insight harvesting

Detects original thinking, contrarian perspectives, and unique frameworks during natural conversation and captures them as permanent, linked notes.

  • Identifies original insights, contrarian perspectives, and novel frameworks
  • Captures as permanent notes with automatic links to related knowledge
  • Works during natural conversation — no manual tagging or filing
  • Distinguishes your unique thinking from common knowledge
Core
02

Multi-layer knowledge retrieval

Three-layer semantic search that retrieves not just what matches, but what's relevant in context using spreading activation.

  • Layer 1: Direct semantic matches to your query
  • Layer 2: First-degree associations from connected notes
  • Layer 3: Extended network traversal via spreading activation
  • Retrieves low-similarity, high-value connections that keyword search misses
Module
03

Cross-domain connection discovery

Finds non-obvious patterns between disparate topics — connections you would never search for but immediately recognize as valuable.

  • Targeted mode: active connections for current research topics
  • Autonomous mode: discovers patterns without prompting
  • Cross-domain linking between unrelated knowledge areas
  • Surfaces insights with explanation of why the connection matters
Module
04

Deep research pipeline

Autonomous three-phase workflow: research with latest papers, extract insights, connect to existing knowledge graph. Runs without intervention.

  • Phase 1: Research across latest papers, articles, and sources
  • Phase 2: Extract 15-25 insights per research session
  • Phase 3: Connect new knowledge to existing graph with links
  • Full reports generated autonomously — review when ready
Module
05

Multi-agent knowledge hub

Serves as the central brain for your agent network. Content, operations, and research agents all query Cornelius and feed learnings back.

  • Ruby (content) queries knowledge base for thought leadership material
  • Corbin (operations) retrieves business context and decision history
  • Specialized agents feed research findings back into the graph
  • Shared intelligence layer that makes every agent smarter
Module

How it works

How Cornelius works

01

Think out loud

Have natural conversations. Cornelius detects and captures your original insights, frameworks, and perspectives automatically.

Step 1
02

Discover patterns

The agent finds connections across your accumulated knowledge — even low-similarity, high-value links that semantic search misses.

Step 2
03

Compound over time

Every insight, connection, and research session makes the knowledge graph denser and more valuable. Your agent network gets smarter.

Step 3

What it delivers

Outcomes, in weeks.

Before → After

Ideas lost in conversations

Permanent, linked notes

Before → After

Research: Hours of reading

15-25 extracted insights per session

Before → After

Manual, obvious connections

Autonomous cross-domain discovery

Before → After

Stateless AI tools

Shared memory across agents

Industry benchmark

Unlike black-box AI systems, Cornelius uses an observable architecture — every note, connection, and reasoning chain is visible and editable in plain markdown. 3,018 notes with 10,565 connections averaging 14.3 links per note.

Where it lands

Built for real work.

01

Thought leadership and writing

CEO or founder building public authorityPull from your knowledge base to create articles, presentations, and frameworks grounded in accumulated insights

02

Strategic decision-making

Executive facing high-stakes decisionsRetrieve past decisions, relevant research, and contextual analysis before making critical calls

03

Multi-agent intelligence

Organization running multiple AI agentsCentral brain for content, operations, and research agents — all benefit from shared knowledge

No lock-in

Built in your stack.

Live in 4-6 weeks. You own the system — sovereign, no vendor lock-in.

  • Obsidian
  • FAISS Vector Search
  • Trinity Platform
  • AI Reasoning (Anthropic Claude)
  • MCP Protocol

Questions

Questions about Cornelius

How is this different from RAG or a regular knowledge base?

Most AI tools use single-layer retrieval — match a query, return results. Cornelius uses spreading activation across three layers, finding connections that are low in direct similarity but high in conceptual value. It discovers patterns you would never think to search for.

How is this different from a note-taking app?

Note apps store what you manually write. Cornelius actively detects original thinking during conversation, captures it automatically, and continuously discovers connections across your entire knowledge base. It's a research partner, not a filing cabinet.

Am I locked into a proprietary platform?

No. All knowledge is stored as plain markdown files in a git-tracked Obsidian vault. You can read, edit, and export everything. If you stop using Cornelius, your knowledge graph remains fully accessible.

Does it actually get smarter over time?

Yes. Every insight, research session, and connection makes the graph denser. More connections mean richer retrieval. Autonomous discovery runs improve as the graph grows — the value compounds with every interaction.

How does multi-agent integration work?

Other agents in your network (Ruby, Corbin, etc.) query Cornelius via MCP protocol for relevant context. They also feed learnings back. Your content agent pulls from your knowledge for articles; your operations agent retrieves decision history.

Can I edit or correct what Cornelius captures?

Absolutely. Every note is a markdown file you can read and edit directly in Obsidian. You can correct facts, update beliefs, delete notes, or restructure connections. Full transparency and control over your knowledge base.

Start here

Bring one workflow.

A 30-minute working call. We’ll map this workflow to an agent stack and tell you honestly whether it’s worth building.

Cornelius — Knowledge & insights

3,000+ notes, 10,500+ connections

AI second brain that captures insights, discovers cross-domain connections, and compounds knowledge.