Blog
Insights on AI Strategy & Implementation
Exploring the intersection of cognitive science and AI, from memory architectures to data ownership and enterprise security.
Featured Articles
Building Production RAG Systems
Key architectural decisions and pitfalls to avoid when scaling retrieval-augmented generation systems. Lessons from 326K+ lines of production code across three distinct RAG architectures.
AI Readiness Checklist: Is Your Company Prepared for Production AI?
An executive-friendly checklist for assessing AI readiness across strategy, data, operations, people, and governance.
OpenClaw-ing Out Your Privacy
Enterprise AI Governance
OpenClaw gained 60,000+ GitHub stars in weeks—then security researchers found hundreds of misconfigured instances leaking API keys and credentials. The real problem? Memory architecture. Here's why AI agent security starts with how memories are stored.
All Articles
Who Owns Your Digital Twin?
Your preferences are fragmented across ChatGPT, Claude, Gemini—each treating you like a stranger. Every new product makes you start over. This isn't a bug. It's a business model. Here's how portable memory changes everything.
Personality Affects Your Memories
Why do some people hold grudges forever while others brush off conflicts? The Big Five personality traits don't just affect behavior—they color the fabric of our recollections. Here's how to build AI that respects individual cognitive wiring.
Where Did I Hear That? Source Monitoring in Memory
You're certain a friend told you something—then realize you heard it on a podcast. Our brains don't store memories with neat labels. Johnson's Source Monitoring Framework explains why, and how AI can avoid the same attribution errors.
An Essay Towards Solving a Problem in the Doctrine of Chances … in AI … from 1763
Beliefs change with evidence. When you hear a rumor, you weigh it against what you already know. Bayes' theorem, 260 years old, turns out to be essential for building AI that handles uncertainty with traceable confidence scores.
Memory & a Game of Telephone from 1932
Memory is not a recording. Bartlett's 1932 experiments showed we fill gaps with patterns based on schemas. Without structured schemas, AI reconstructed memories diverge from stored facts. Here's how ontology grounds retrieval.
Short-term Memory Beyond the Context Window
Working memory holds information for seconds to minutes. Baddeley's multi-component model—central executive, phonological loop, visuospatial sketchpad—maps directly to how AI should manage conversation context without overloading.
Even AI Needs Sleep
During sleep, your brain reorganizes information into long-term memory through consolidation. Periodic offline processing applies the same approach to AI: clustering experiences, extracting insights, pruning noise.
Tulving's Memory Model and What It Means for Digital Twins
Tulving's 1972 framework split memory into two systems: episodic for autobiographical experiences, semantic for general facts. This distinction maps directly to building AI that maintains useful context about individual users.
Why AI Memory Is the Problem Nobody's Solving
AI products forget what you told them yesterday, give generic advice, and require constant hand-holding. The reason is simpler than most people think: they have no real memory architecture.
Have Questions About AI Implementation?
Let's discuss your specific challenges and how I can help.
Get in Touch