Research & practice
How I finish real work with AI agents
I started with the result I wanted, worked through small revisions and turned it into something I could use. Here is the approach I use for documents, software and home systems.
The problem
I knew the result I wanted, but not every technique needed to build it.
My approach
I described the sources and behavior I wanted, then tried and revised small results with the agent.
What changed
My ideas became files and systems I could open, run and keep improving.
BUILD DETAILS
Build details & ideas to reuse
01The problem: I knew what I wanted, but not how to build it
I did not know every tool or technique before making reports, slides, teaching tools or systems for my home. I started by describing the result: the file I needed, or what I wanted a device to do. Choosing the tools came after that.
02The process: I worked from small results and revised them
I gave the agent source material, photos and a description of the behavior I wanted, then started with a small result. For documents, I checked the text, tables and sources. For software and devices, I tried the task myself. I pointed out what did not match and revised it with the agent.
03What changed: ideas became things I could use and keep improving
I came away with documents and code I could open, run and revise. When I wanted to change part of a presentation or connect another device at home, I could build on what was already there. Each Lab case shares both the result and how I made it.
04How it works: I keep the source files and the setup together
For documents, I keep editable originals with the reference material. For software, I keep the code, configuration and instructions for running it. That makes later changes easier to trace. Before sharing the work, I remove private information and distinguish actual results from explanatory images.