Breaking complex coding problems into smaller chunks gets better AI results than fancy prompts. After 30+ years coding, I’ve learned the real skill is decomposition, not prompt engineering.
The story
I’ve been coding for over 30 years, working with AI for the last eight, and I’ve noticed something: the developers who get the most out of AI assistants aren’t the ones with the fanciest prompts. They’re the ones who break problems down properly.
Most people throw entire projects at their AI tools and wonder why the output is mediocre. The real skill is in decomposition – taking a complex problem and splitting it into chunks an AI can actually handle well.
When I start a project, I don’t jump straight to prompting. I map out the problem first. What are the natural breakpoints? Where are the dependencies? I sketch it out, visualize how pieces connect, and identify which parts are good candidates for AI assistance.
Then I modularize everything. Each subtask gets clear boundaries and specific goals. Not so big that the AI gets lost, not so small that I’m spending all my time orchestrating.
I use techniques like chain-of-logic prompting, where I break reasoning into explicit steps. Or plan-and-solve, where the AI maps out its approach before executing. For math-heavy work, I’ll use program-of-thoughts – expressing logic as code to keep calculation separate from reasoning.
The key skills that matter: writing clear prompts, keeping components modular and testable, managing context carefully, and validating each chunk’s output before moving to the next.
When I work with multiple AI agents, I treat them like team members. Each gets a distinct role with clear boundaries.
I never expect the first attempt to be perfect. I test each piece separately, refine based on what comes b
Recent Comments