Chapter 12: The Knowledge Spiral
AgentSpek - A Beginner's Companion to the AI Frontier
Working with AI teaches you new things, and it keeps showing you how much of what you 'know' is shallow or simply wrong. It happens at a pace that is disorienting, and it does not stop.
A new tool does not solve all the problems. It frees us up for other ones, and the other ones turn out to be harder, and more interesting, and more human.
Wrong About Everything
You think you know something. Then you ask the AI for help and it casually corrects an assumption you have held for years. Your first instinct is that the AI is wrong. You check the documentation. And then you realize you have been working with an outdated mental model, one that was “good enough” in practice but technically incorrect.
This keeps happening, constantly. The framework you thought you understood has edge cases you never ran into. The pattern you thought was optimal has better alternatives you never heard of, and the architecture you thought was standard practice moved on while you kept building the old way.
Working with AI teaches you new things, and it shows you how much of what you “know” is shallow or simply wrong. It happens at a pace that is disorienting.
Unlearning
AI speeds up learning, and it speeds up unlearning more, the painful and necessary work of finding out that things you thought you knew were incomplete or out of date or wrong.
The old way, you found the gaps gradually. A conference talk, a code review, a production bug that teaches you an edge case you had overlooked. Those moments were spaced out over months or years, with time between them to fit the new understanding in with what you already had.
With AI it comes all at once. In a single afternoon working with Claude on a Neo4j integration for my blog, I found out that my understanding of graph database indexing was superficial, that my approach to Cypher query optimization was inefficient, and that my mental model of relationship traversal performance came from relational database assumptions that did not apply.
It was less new information on top of old knowledge than the uncomfortable realization that what I knew was standing on shaky ground. The AI filled gaps, and it showed me gaps I did not know I had.
Expertise used to come with a kind of confidence. You knew what you knew and you knew what you did not. Working with AI opens a third category, the things you thought you knew and did not, and that category grows faster than the other two. Epistemic vertigo.
The Mirror
The AI turns out to be a mirror for my own thinking. Explaining a problem to it, giving it the context, going back and forth over its suggestions, I end up looking at my own mental models with unusual clarity.
A lot of programming is intuitive. I “know” the right approach without ever saying why, I follow the patterns that feel right, I make architectural decisions out of experience. It works, and it leaves the knowledge tacit.
Working with the AI pulls the tacit out into the open. To get good results I have to say what I want and why. The constraints I take for granted have to be explained, and the assumptions that usually stay buried have to come up.
Through months of AI conversations I discovered I have a strong bias toward stateless architectures, consistently underestimate the importance of error handling in initial implementations, and tend to optimize for developer experience over runtime performance. I did not get those from self-reflection, or from working with human colleagues. The AI needs the context spelled out, and spelling it out forced a self-examination that normal programming work never asked of me.
You’ve read the opening sections of this chapter. The full chapter (Distributed Expertise, Collaborative Creation, Infinite Information, Temporal Collapse, Questions Over Answers, Cognitive Empathy, Meta-Learning, Wisdom, The Future of Knowing) continues in the book.
Chapter 12 of 18 in Chapter 12: The Knowledge Spiral