Anthropic just made me sit an exam on how to TEACH AI, not how to use it.
I already have the certificates for Claude 101 and Claude Code, the builder side, learning how the tool actually works under the hood. Right now I'm on Teaching AI Fluency, a different track entirely, built with Prof. Joseph Feller (University College Cork) and Prof.Rick Dakan (Ringling College) on their AI Fluency Framework.

This course grades you on whether you can design assignments that make people both build AI fluency and prove they built it, while keeping the marking load manageable as AI-assisted work multiplies.

That's not a tool question. That's a teaching methodology question, and it made me name something about my own sessions I'd never said out loud: I already build them this way instinctively. I don't measure a training by whether people were impressed during the demo. I measure it by whether they can still do the thing a week later, unsupervised.

In conclusion (even if it seems obvious, this will change who you engage to train your team or yourself):

- Knowing Claude well and being able to teach Claude well are two different skills
- A demo makes people say "wow." A real training makes people produce something and defend it
- If a session ends with applause instead of graded output, it was entertainment, not training.

Here's the part that should worry anyone assigning "AI training" internally: the person who's fastest with the tool in the office is rarely the person who can design that second part. Most companies hand the fastest person a slide deck and call it done. Three weeks later, nobody remembers the session because nobody had to prove anything.

This is exactly why I went and got graded on the teaching, not just the tool. It's the part of AI adoption training that's invisible until it's missing.

Who's actually running the AI training at your company right now, the fastest person with the tool, or someone with an actual method behind it? 🤨