AI Journey Chronicles: a chapter-by-chapter record of learning to work with AI on real business software.
AI drafts that invented my backstory, emoji-filled posts, and too much "I": how a written voice guide keeps AI writing sounding like me.
From copy-pasting AI code to a coding agent that edits files directly: what got faster, what didn't, and why pricing by the hour stopped making sense.
Building with AI meant copy-pasting code through my clipboard and FTP, like Dreamweaver days. The rules that came out of it, and why I moved to git.
Two AI chats gave opposite verdicts on the same question. Why AI research is a draft, and the prompt rules that stopped it inventing facts.
A second AI pass that agreed with the first instead of reading the image, and a model that claimed to count but didn't: building checks around AI output.
How AI made JavaScript inside FileMaker web viewers practical: a batch engine, a race condition, and the old-fashioned fix that finally worked.
ChatGPT invented a FileMaker function that never existed. What AI-written code gets wrong, why the runtime matters, and how to test before you trust it.
The year AI answers stopped landing in a chat window and started feeding software: prompts as specs, JSON output, and code built by description.
An AI coding agent overhauled this website in one evening: staging first, backups, dry runs and human approval for every live change.
GPT-4 for social posts, Stable Diffusion characters, RunwayML animation, and the first time code entered the chat: what a year of AI content taught me.
