Skip to content
← Journal

State of AI literacy: what learners need next

Why model fluency is not enough for responsible use.

Updated 8/23/2026

The gap
Lots of learners can get a model to write a paragraph. Far fewer can tell you when that paragraph is a hallucination, whose data it was trained on, or why a district should not paste IEPs into a consumer chatbot. Model fluency is spreading faster than judgment. That is the literacy problem.
What 'enough' looks like
Name the tool class (calculator vs. search vs. generator). Write a prompt with a check. Verify high-stakes claims. Refuse to submit generated work as your own. Know that safety is operations plus design, not a badge on a homepage. WonderMind stages that sequence from age four through Pioneer. The public learn pages that are still outlines are marked noindex until they are real guides.
What schools are buying by accident
If the pitch is 'SSO, Clever, SCORM, API' and the repo does not have those toggles, you bought a slide. If the pitch is join codes, objectives, no child DMs, homework refusal, and a help center that argues with marketing — that is a product you can evaluate. Demand the second. Ignore the first until a sandbox IdP exists.
What to do this week
Run one experiment from the journal. Read /safety. Try the demo family. Do not put a six-year-old's real record in until legal says the word draft is gone.