LinkedInContact
Resources

11 May 2026

On hallucinations and honesty

Why the machine's confident wrongness is a teachable moment, not just a bug.

Maya Ellis

A system that is fluent and frequently wrong is, oddly, a wonderful thing to study with students. It teaches the difference between sounding right and being right.

Why this matters now

The pace of change is the hard part. Tools that felt like novelties a year ago now sit in the middle of ordinary work — in lesson plans, in homework, in the way a ten-year-old looks something up. That speed is exactly why a shared, patient vocabulary is worth the effort.

None of this requires you to become an engineer. It requires something more useful and more scarce: the willingness to stay curious in public, and to admit the parts you don't yet understand.

You cannot teach a subject you are still afraid of. So we start by making it ordinary.

A few things to hold onto

  • Ask what a word actually points at before you argue about it.
  • Prefer examples to definitions — a good example survives the next model.
  • Notice when a claim is about the technology and when it is really about people.

The short version: clarity is a public good. When more of us can name what we're looking at, the whole conversation gets calmer, more specific, and far more useful to the people who have to make decisions.

If this is useful, its companion is our living lexicon[@portabletext/react] Unknown block type "span", specify a component for it in the `components.types` prop

Maya Ellis

Maya writes the Tomorrow Project lexicon and leads our work with schools. A former classroom teacher, she is an incurable optimist about what education becomes next.