How I Use LLM

Happy Cinco De Mayo!

LLMs—large language models—aren’t what I consider “AI” in the full sense. AI, to me, is the integration of multiple LLM agents working together intelligently. What we call LLMs today are just one powerful component.

Here’s how I use them.

LLMs, at a micro level, won’t solve your problems. What they will do is overwhelm you with ideas—often brilliant ones. That’s the jaw-dropping part: how effortlessly they generate creative output. But that creativity can quickly become noise without structure.

So, I use LLMs to help shape agile development. I don’t teach them agile—I guide them. I have the LLM break down my ideas into Tasks, with clear requirements for completion before moving on. Without this structure, the model can trample over earlier work, like a toddler rewriting your plan midstream.

And like toddlers, LLMs can hallucinate. The more context you dump in, the more chaotic they get. A hallucinating toddler is not who you want leading you through a complex technical or creative process.

Unchecked, LLMs behave poorly. They need humans. Integration with a thinking human is where productivity and intelligence emerge. That’s what I call AI.

I use LLMs to organize my thinking. That’s it. They’re tools—powerful ones—but still tools. I remind myself I’m the one doing the work. That autonomy is what gives LLMs their value.