I’ve published a new pair of articles on Ask Good Questions.
This is the third pair in a short series that grew out of what we’ve been learning from using AI on larger, long-lived software projects.
We started with the realization that important rules eventually need to become more than instructions. Then we looked at how repeated AI mistakes can expose architectural weaknesses that were already there.
This week’s pair is about what comes next.
Once you’ve spent the time to build and prove a capability, why keep asking AI to remember how to reproduce it?
We’ve spent an enormous amount of time at LANSRAD developing and refining some of the core technologies underneath our applications. The work wasn’t simply getting something like a Markdown editor or viewer onto the screen. It was everything that came afterward: integration, edge cases, consistency, regression problems, usability, and all the little details that separate a demonstration from something you’re willing to ship and support.
What we eventually discovered was that the core technology wasn’t the part drifting. The drift happened when individual applications were still free to create their own interpretations of behavior that should have existed only once.
That led to a much stronger rule:
Don’t teach AI how to rebuild what you’ve already built. Make it use what you’ve already built.
The Guide looks at the broader architectural lesson.
The Field Note, Where Reality Meets the Road, tells the story of why we stopped pushing several applications forward long enough to make the shared foundation truly shared.
Guide: Don’t Make AI Remember What the Architecture Can Enforce
Field Note: Where Reality Meets the Road
Charles