Here’s what’s actually happening on X.

Representative of Ethan's classroom threads — using Claude to run simulated negotiations and case discussions, with the model playing both sides and reflecting on the student's moves.

“The tutor isn't replacing the professor. It's replacing the empty chair in the study group.”

You are playing <role> in a <scenario> with the user. Stay in character for up to 10 turns. Your objective: <counterpart goal>. Do not hint at the 'right' answer. After turn 10 OR if the user types 'end,' drop character and produce a written reflection: (1) three moves the user made that advanced their goal, (2) two they should reconsider, (3) one alternative approach. Cite specific turn numbers.

Representative of instructional-coach threads — teachers using Claude to give first-pass essay feedback so they can spend their 1:1 time on the harder questions.

“The AI handles mechanics. I handle meaning. It's the best team I've ever taught with.”

Representative of history-teacher posts — using Claude to translate primary sources to modern English, generate discussion prompts, and check student sources for hallucinations.

“I teach source analysis better because the AI gets it wrong just often enough to be a teaching moment.”

Representative of Davy's threads on AI adoption in African classrooms — offline-capable local models via Ollama serving students without reliable internet.

“AI equality isn't a frontier-model problem. It's a local-model deployment problem.”

Representative of UK teacher threads on AI lesson planning — retrieval quizzes, starter activities, exit tickets all generated from a unit specification.

“Planning a full unit used to take 6 hours. It's a 45-minute conversation with Claude now.”

Representative of district-level ed-tech posts — rolling out AI policy, staff PD, and student acceptable-use agreements across a district of 20+ schools.

“A good AI policy isn't a ban list. It's a value statement plus a practice guide.”

Representative of K-8 teacher threads on X — using Claude to generate differentiated reading passages (same content, 3 reading levels) in a single prompt.

“Differentiation used to be my Sunday. Now it's a Monday morning prompt.”

Representative of Ethan's academic-research threads — using Claude and Deep Research to do literature reviews that would have taken a grad student weeks, in hours.

“The jagged frontier is real. AI is brilliant at some parts of research and useless at others. Learning which is which is the job.”