Insights
Notes on agents, architecture, and alignment
How Daiven is built, why it's built that way, and what we've learned about AI agents along the way.
Published by category
- AI04
- Architecture02
- Psychology01
The archive 07 articles
Architecture
AI Agent Memory for a Self, Not a Task
Most AI agent memory extracts facts, keeps the newest and expires the rest. Right for a task. For a record of a person's values, each default reverses.
September 29, 2026Przemek Czerpiński8 min readRead →
AI
Why Does AI Agree With Everything You Say?
Why does AI agree with everything you say? The training, the psychology, and the product incentive behind AI sycophancy — and what the agreement costs you.
September 15, 2026Przemek Czerpiński9 min readRead →
AI
AI Hallucination: Why Models Make Things Up About You
AI hallucination about a person works differently from an invented citation. Nothing external corrects it. What causes it, and what removes the cause.
September 7, 2026Przemek Czerpiński10 min readRead →
Architecture
AI Memory: Why Assistants Learn to Agree With You
Most AI memory is one file that only grows. What that costs, why an assistant that learns from you drifts toward agreeing, and how Daiven fences it off.
August 27, 2026Przemek Czerpiński10 min readRead →
Psychology
The Psychology of Staying Aligned With Yourself
You already know what you value. Why staying aligned with it is harder than it sounds, where the other voices came from, and why AI settles the question wrong.
August 12, 2026Przemek Czerpiński7 min readRead →
AI
What Is an AI Agent? Meaning, Skills, and Limits
An AI agent is a chatbot that acts. What that changes, what agent skills are, and the five questions worth asking before you trust one with anything.
August 12, 2026Przemek Czerpiński8 min readRead →