Blog

I didn't finish earlier versions of Open Scaffold Templates. This one, I think, is the one that will be worth building.

I'm still working on it and probably will be for a while longer. This time, though, I'm looking at the process differently: Looking for ways to remove rules, not add them. Looking for better ways to describe outcomes and options to a user without telling the user or the AI agent how to achieve that outcome.

Claude drafted instructions to be given to new users. I read the first draft. And three times I was disappointed to find the same problem.

• Where the draft, talking about table names, should have said prefix, it said marker.
• Where it should have talked about queries and procedures, it said pieces.
• Where it should have said record, it said row.

That's almost as far as I got before I lost my cool.

...

The painful thing is that those choices followed my "plain language" rule exactly.

Anything you write to an AI in the imperative will be obeyed, silently and immediately. That's a feature. It's also why every "just do X" sentence in your documentation is a trap waiting to surprise you when you least expect it.

You didn't become the lead in your team because you know more than anyone else in your shop. You became the lead because you actually know less than the AI agents you manage. And now you are learning on the job what the task of managing an AI agent really involves.

Sometimes an AI agent proposes something plausible that turns out to be wrong .... I correct it. It tries again. The next pass is better, and it's better for a reason I can articulate, because the correction forced the reason into the open. That loop is the work.

It's the same loop I've been running on myself for thirty years, just faster.

I pay a flat monthly fee for my AI coding assistant, so a session which my tracking tool says “cost” $265 actually costs me nothing on any invoice. So why track it? The dollar figure may be imaginary but the thing it measures is not. On a subscription you spend a real, limited resource on every turn, and most of it goes just to carry the conversation forward.

When you ask an AI assistant to build database tables, it designs to its own defaults, not yours. Open Template Scaffolds flips the order: the AI reads a template and your standards, shows you a design to approve, and builds nothing until you say yes. It's free, open-source, and public now on GitHub.

I built DevPulse to answer a simple question: is an AI assistant actually cost-effective? One real session -- a supposedly minor bug fix -- gave me an answer I didn't expect, and along the way exposed something uncomfortable about my own discipline, not just the AI's.

Not that long ago, I was convinced AI was a solution in search of a problem. Lately I think I'm beginning to figure out just what that problem might be. And more importantly, I'm focusing more on how use an AI to produce real, legitimate solutions. Not unsurprisingly, it's not at all what I thought it would be.

Sometimes the light's all shinin' on me

Other times I can barely see

Lately it occurs to me

What a long, strange trip it's been.

-- The Grateful Dead, Truckin', 1970