Gary, a British ESL teacher now based in Cambodia, reflects on a decade of teaching, why AI can't replace human connection, and what kids chasing influencer fame are missing.
A decade into teaching English as a second language, Gary has seen the industry shift in ways that surprised him. Students who once crammed into training schools six days a week are now being taken by their parents on weekend trips to Vietnam or Indonesia. Saturday mornings, once packed back-to-back, now sometimes hold a single class. He takes that as a good sign.
"I think students are now giving a little bit more time to not relax, but do some other things," he says. It's a small change, but for someone who watched children studying on the back of motorbikes in China, falling asleep over their books, it feels significant.
Gary spent years teaching in China before arriving in Cambodia to cover classes for a friend who was unwell. The recovery took longer than expected, and he's still there, running a summer camp with roughly thirteen kids split across two grade levels. The principal has started asking how long he plans to stay.
Teaching in China used to offer room for creativity. Gary remembers being able to choose how he taught, to go off-script when a lesson called for it. That flexibility has narrowed considerably. Friends still working there tell him teachers are now handed a textbook and told to follow it precisely, with little room to deviate.
Cambodia, by contrast, lets him teach Minecraft Education. One session, the principal sat down at the keyboard thinking it was just a game kids waste time on. Gary walked him through building a hospital, planting a rooftop garden to feed the patients, hiring workers from the city to harvest the crops.
Build a hospital. What do you need to put in it? You can grow vegetables on the roof to feed the patients.
The principal came around. Gary sees that kind of project-based thinking as exactly what kids need more of, especially when the alternative is three and a half hours of YouTube for every twenty minutes of homework.
He asked his students to keep an honest log for a week: homework time on one side, screen time on the other. One boy: twenty minutes of homework, three hours thirty of YouTube. The numbers didn't shock Gary so much as clarify something he already suspected.
Ask a classroom of kids what they want to be and a fair number will say influencer. Gary pushes back, not with dismissal, but with a genuine question: why would anyone watch you?
At some point, people are going to get truly bored with watching people buy ice cream.
The kids who became the first generation of successful influencers got there because one video did something genuinely different. That edge is harder to find now, and most kids chasing it haven't thought through what they'd actually bring. Gary's alternative pitch, doctor, police officer, firefighter, lands with a predictable response: too much hard work. He understands why. When you can see someone playing games online and earning money from it, the gap between effort and reward looks very different than it does from the other side.
The students who do well in his classes tend to come from families where parents already have a clear-eyed view of screen time. Not zero-screen families, Gary teaches on a screen himself, but families where the hours are counted and questioned.
The English learning industry is having the AI conversation loudly right now. You can open ChatGPT, describe your target IELTS level, and get a daily study plan tailored to your current vocabulary and available hours. It works, in a narrow sense.
"It's not the same as this. It's not real." — Gary
Gary's student, who is on the other end of this conversation, tried learning Japanese through AI. Two or three sessions in, she stopped. It felt lifeless. Gary's own experience with AI tools lands differently: he used one to build an interactive grammar workbook for his students, with four deliberate mistakes baked in. AI as a preparation tool, not a replacement teacher.
The thing Gary keeps coming back to is accountability. An AI monitors his lessons through a small robot in the room, green eyes blinking, sending reports with suggestions. He doesn't mind the feedback. What concerns him is the possibility that the observation is also a form of replication: that the system is learning what he does well enough to do it without him. He doubts it, mostly because, as he puts it, it would have to learn from him and he's not confident enough in his own teaching to think that's much of a threat.
The real limit of AI in teaching isn't technical accuracy or lesson planning. It's that it hasn't lived anywhere. It hasn't seen a man walking a camel the wrong way down a road in Fujian, hasn't managed a student so angry he threw furniture, hasn't spent a Saturday morning with one class instead of twelve and found the quiet useful. Those experiences are what make a conversation feel like more than an exchange of information.
One of Gary's first moves in a new classroom is a list: what's acceptable, what isn't, what happens when you cross the line. Day one, every year. A student at the Cambodia camp arrived with a reputation for throwing chairs when he got frustrated. Gary's read on it was simple: if you don't set expectations, you should expect exactly that behavior.
The lever he found with that student was Minecraft Education. If you want to build the hospital, you don't shout at me. You don't throw things. The boy's anger, according to Gary, dropped off quickly once he had something to work toward and a clear cost for blowing it.
If you want to play Minecraft with me, why would I let you when you throw a chair?
Gary relates to that student more than he lets on. He describes his own childhood self as a horror, a terror, the kind of kid his parents didn't quite know what to do with. The kids who push hardest usually want something the most. Finding out what that thing is, and making it conditional on decent behavior, is less a discipline strategy than a negotiation between two people who both want the lesson to be worth showing up for.
That negotiation, messy and specific and built on knowing the person in front of you, is what he means when he says AI can't do what he does. Not because the technology is primitive. Because the relationship isn't transferable.
