Speed Is Useful. Reliability Is Required.

When my kids were young, I used to read them a Goofy book.

Goofy was in a race. His line, over and over: slow and steady, steady and slow, that's the way we always go. Donald had the flashy new car. He kept taking shortcuts. Goofy just kept driving.

Same lesson as Aesop's tortoise and the hare. Two thousand years old and we apparently keep needing to hear it.

I think about that a lot lately.

Every few weeks there is a new reveal. First it was prompts. Then agents. Then loops. Then the next thing, before the last thing finished proving itself. Every few months there's a new answer to everything. Then the next one arrives.

I watched that pattern from inside it, not outside. I use Claude, GPT, and Grok every day. AI has made my work significantly faster. I love the innovation. I love the speed.

Every hype cycle pulls people in the same direction at the same time. Nobody stops to ask whether the direction is right. What I can't ignore is that the reliability problem hasn't been solved by any of it.

I build training for regulated industries. Compliance courses. Legal education. State licensing curriculum. Somebody signs the deliverable. That's different from a demo on X.

My process barely changed through all the hype cycles. I still pass drafts between models. I still verify important claims. I still make the final call. That's the thing processes do that tools don't. They survive tool changes.

Last year I worked on a state licensing project in Arizona. I didn't use a dynamic AI simulation. I used Bloom-aligned questions, cross-model review, and curated feedback written with AI assistance and verified before it went anywhere. The learners got better feedback than most live AI systems produce. Because a human was running the loop.

I don't distrust AI. I distrust unverified AI.

The tools change. The need for judgment doesn't.

Donald wasn't wrong to want a faster car. I'm not against the faster car. I'm driving one. The problem wasn't the speed. The problem was that he never seriously considered he might lose. He had the newest thing. He thought that was enough.

Goofy won.

Speed is useful. Reliability is required.

—

Mariane McLucas is a senior instructional designer and solo consultant at Module Makers. She is developing a training program for professionals who want to use AI reliably in their real professional workflows. Details at modulemakers.com.