Lately, I’ve been thinking about how working with AI tools like ChatGPT or Cursor has subtly—but fundamentally—changed the way I approach building software. It’s not just that they make things faster or more efficient. It’s that they’ve introduced a new kind of workflow. A new way of thinking.
If I had to describe it in a single word, it would be: fractal.
When you work with AI tools—whether you’re building an app, writing a blog post, or exploring an abstract idea—the process takes on a kind of recursive, branching shape. You start with a seed, a root prompt, and from there, ideas begin to bloom outward. The more you engage, the more the structure unfolds.
It’s not linear. It’s not hierarchical. It’s not even fully predictable.
It’s exploratory—and it feels a lot like zooming into a fractal.
🌱 The Root Prompt
Every AI interaction begins with an initial prompt—a question, a goal, a hypothesis. That prompt is the root of the tree. From there, each follow-up becomes a branch, and each clarification, refinement, or tangent becomes a branch of that branch.
Ask a general question, and you’ll get a broad response. Ask the AI to zoom in on a specific piece of that response, and you’re down a new path. Each conversation becomes a kind of living map—a tree that grows and forks in real time.
Sometimes the branching is deliberate: “Let’s go deeper on this module.” Sometimes it’s intuitive: “Actually, what does that word mean again?” Sometimes it’s accidental: “Wait… what API?”
And sometimes—importantly—you realize you need to pop back up to a higher branch to rethink your direction.
🔁 Recursion as Workflow
This recursive, zoomable style of working isn’t unique to AI. It mirrors how experienced developers naturally think: break down the problem, isolate the pieces, drill into the tricky parts, recombine.
But what AI does is amplify this. It lets you move up and down the abstraction stack effortlessly. You can ask for a high-level architecture one moment, and the next, you’re generating a regex pattern or rewriting a single function with a performance constraint.
It’s like working with a collaborator who’s always ready to follow your train of thought—whether you’re zooming way out or all the way in.
🍂 Dead Ends and False Branches
Of course, not every path is fruitful. One thing I’ve noticed—and this is important—is that AI agents are excellent improvisers. Sometimes too good.
They’ll confidently suggest solutions that don’t exist. They’ll generate fake API endpoints. They’ll promise functionality that sounds reasonable but turns out to be completely unsupported.
For example, I once asked the AI how to pull metadata for my Audible library. It confidently told me about an “Audible API” that doesn’t actually exist. Only later did I realize that the only way to get that data is through scraping—because Audible is locked down tighter than a drum.
In the fractal metaphor, this is a dying branch. It looked promising. You followed it for a bit. But eventually, you had to prune it and return to a previous fork in the road.
This isn’t a flaw—it’s part of the process. It’s what exploration looks like.
🧠 Conversations as Code Trees
This pattern—branching, backtracking, diving deep—isn’t limited to software. I’ve experienced it while researching historical topics, brainstorming product ideas, even just reflecting on personal experiences.
It’s made me realize: every AI conversation is a fractal.
You’re not just generating content or solving a task. You’re growing an idea-tree, in real time, with a partner who helps you map the terrain.
Sometimes, you zoom in so far you forget where you started. Other times, you pop back up and realize you’re at a different tree entirely. Either way, the shape of the process is always recursive, always expanding, always alive.
🌳 Toward a Fractal Methodology
I’m starting to think this isn’t just a metaphor—it’s a methodology. A new way to work.
Not waterfall. Not agile. Something more organic. More dynamic. More emergent.
A fractal workflow means:
- You can start anywhere.
- You explore recursively.
- You prune and re-branch as needed.
- You stay fluid, rather than fixed.
And most importantly, you let structure emerge through dialogue, not upfront planning.
✨ Closing Thoughts
There’s something beautiful—and strangely natural—about this process. It’s not about commanding the machine or automating the work away. It’s about thinking with the machine. Co-creating. Mapping possibilities. Exploring mental terrain you didn’t even know was there.
And if that sounds a little trippy… well, maybe it is.
But more on that another time.