How three educators are actually learning AI

"AI puts educators in the design seat, instead of waiting for someone else's tool to fit our context."

Working with Punya Mishra taught her that there is no such thing as educational technology. Every tool was designed to make a profit. PowerPoint, Canvas, all of it. AI lowers the cost barrier so educators can build the tools their classrooms actually need, instead of bending borrowed ones.

"My students are my why. When AI came about, I saw multimodality."

As a math teacher of English language learners, he could not always tell what his students were doing in transformations or dilations through text alone. AI changed what formative assessment can look like. The orientation is to every single student, not to his own workload.

"My why is to figure out the potential here. But I also want to know where the dangers are."

When ChatGPT launched, she dug in instead of banning. She wrote a chapter on using AI to help, not hinder, learning. The driving question is one she came to from teaching: where do students need productive struggle, and what can they offload?

"When AI fails, students realize they need a human in the loop. That's where the friction lives."

She does not try to put friction back into AI. She watches for the moments AI breaks down and uses those as the teaching surface. The job, when working with faculty, is to make sure the design does not erase the struggle students need to grow.

"If your effort over output ratio is close to one, learning is happening."

AI is designed not to have friction. The educator's job is to bring it back. He uses a working ratio. Effort divided by output. Close to one means students are doing the cognitive work. Far off in either direction means there is a problem on either side of the tool.

"Memory is the residue of thought. We don't absorb. We think, and that's what we remember."

She borrows Daniel Willingham's line as the test. The question is not whether AI is in the room. The question is whether students are still doing the thinking. If AI is helping them think more deeply about content, fine. If it is replacing the thinking, the residue is gone.

"I've made about 30 apps in two months. The learning is in asking other people."

She has taken courses and certificates, and found most weren't built for educators. The shift was to project-based learning. Build something real. Then ask other people which tool, what it costs, what you give up in privacy. The peer network is the curriculum.

"Paid, opt-in, with a teacher on special assignment as facilitator. Both the structure and the near-peer matter."

Irvine Unified runs an AI Pioneers PLC. Five times a year. Structured peer collaboration plus a thought partner who is one step ahead. He credits both. The structure makes the time exist. The near-peer makes the learning specific.

"I didn't sign up for classes. I set myself challenges and tried to break things, safely."

She skipped the formal courses on the theory that there are no settled best practices to teach yet. Her version of learning is annotating chatbot conversations, writing chapters, breaking platforms in low-stakes places, and leaning hard on cross-institution peer networks.

"When someone benefits from a technology, someone else is usually being disenfranchised by the same one."

Working alongside an informatics department changed her frame. Reading Safiya Noble and Ruha Benjamin made the questions about each tool sharper. Tokens, privacy layers, environmental cost, who pays. Her stance became: never jump in. Understand what you want to do first.

"It took me three years to come to this point: not every teacher needs to use AI."

His evolution was past evangelism. Most opposition to AI in schools is opposition to ChatGPT specifically. The question for educators is whether they have a problem of practice that AI is part of the answer to. If they don't, the answer is not to push the tool. It is to ask a better question.

"I will never use AI to create video. The energy use is exponential as the video gets longer."

She found what-uses-more.com, which compares the energy use of AI tasks against familiar baselines. The shift was technical and ethical at once. Now she always starts with the simplest, lowest-tier tool and only goes deeper if she needs to. The tools are not going anywhere. The question is how to use them responsibly.