I expected prompt tricks to matter more. The classroom evidence changed my mind: the real AI skill for K-12 teachers is not clever prompting. It is building a clear rubric, the scoring guide the model is supposed to follow. [C002]
If you only look at the surface hype, you spend time, budget, and attention on the wrong thing. A post is worth your time not by how many features it lists, but by whether it changes your next decision.
The clearest example here is small but useful. In a workshop with 25 K-12 teachers, AI-generated rubrics helped as first drafts and exposed fuzzy standards, but they often landed at the wrong grade level or drifted away from the lesson target. Teachers still had to rewrite them.
A second study pointed the same way. When the model got teacher-written rubrics, class materials, and sample essays, feedback became more specific and teachers saved time. But the scores still needed human checks. That is support, not autopilot.
So if you care about Helping K–12 educators build practical AI skills [C001], start with one assignment, one rubric, and one feedback style before you start teaching prompt tricks. Share this with the person planning AI training in schools.