After 30 Years Coding I Cracked AI

Published 2025-11-14 06-03

Summary

After 30 years of coding, I learned the secret to getting useful work from AI: break problems into small chunks instead of dumping everything into one prompt.

The story

I’ve been coding for over 30 years, and here’s what I just figured out: the best way to get useful work out of AI is to break your problem into chunks.

Sounds simple, right? But most people don’t do it. They dump their entire mess of a problem into ChatGPT and get frustrated when the output is vague or wrong.

I spent the last 8 years building AI solutions, and I finally see why decomposition – breaking big problems into smaller pieces – works so well. It’s not just good practice. It’s the practice.

When I give AI a smaller, focused task, it performs way better. The accuracy goes up because there’s less room for the model to drift or hallucinate. I can test each piece independently, catch errors early, and actually understand where things go sideways.

Here’s my process now:

I start by identifying the subproblems. What are the distinct tasks hiding inside this bigger challenge? I list them out – either on my own or by asking the AI to help me break it down.

Then I tackle each one separately. One prompt, one function, one module at a time. No kitchen-sink requests.

Once I have solutions for the chunks, I stitch them together. Clean interfaces between modules. Clear inputs and outputs. Test each part before integrating.

For complex stuff, I use techniques like Plan-and-Solve Prompting – where I have the AI lay out a plan first – or Tree-of-Thought, where it explores multiple solution paths.

Example: I needed an AI-powered web scraper recently. Instead of one giant prompt, I broke it into scraping, parsing, data storage, and report generation. Each got its own prompt. E

For more about Skills for making the most of AI, visit
https://linkedin.com/in/scottermonkey.

[This post is generated by Creative Robot]. Designed and built by Scott Howard Swain.

Keywords: DecompositionPrompting, AI prompt engineering, problem decomposition, coding productivity