What is Cognition Engines?

Cognition Engines is a decision intelligence platform for AI agents. It helps agents make better decisions by learning from past choices.

The Problem

AI agents make hundreds of decisions per session - architecture choices, tool selections, process changes, bug fix approaches. But they:

• Don't learn from past decisions - each session starts fresh
• Can't search their history - "what worked last time?" has no answer
• Lack guardrails - nothing prevents repeating known mistakes
• Have no calibration - confidence estimates are unchecked guesses

The Solution

Cognition Engines provides three capabilities:

1. Accelerators

Cross-agent learning via semantic decision search. Before making a new decision, query the corpus to find similar past decisions and their outcomes.

cstp.py query "handling transient API failures" --bridge-side function --top 5

2. Guardrails

Policy enforcement that prevents violations before they occur. Define rules like "no production changes without code review" and the system enforces them automatically.

cstp.py check -d "deploy to production" -s high -f 0.85

3. Auto-Capture

Every decision automatically gets:

• Deliberation traces - which queries and checks preceded this decision
• Bridge-definitions - structure (what it looks like) + function (what it solves)
• Related decisions - linked predecessors from pre-decision queries

Theoretical Foundation

Cognition Engines is inspired by Marvin Minsky's Society of Mind:

• Ch 12 - Bridge-Definitions: Describe concepts by both form and purpose
• Ch 18 - Parallel Bundles: Seek multiple independent reasons, not one serial chain
• Ch 27 - Censors: Proactive warnings that intercept before mistakes, not just reactive blocks
• Ch 28 - Mental Currencies: Confidence scores that preserve reasoning structure

What It's Not

• Not a database - it's an intelligence layer on top of storage
• Not an LLM - it doesn't generate decisions, it helps agents make better ones
• Not a logging system - decisions are indexed, searchable, and connected