I am not the sole source of knowledge in the world, nor the classroom. I am a guide, a mentor, and a facilitator. Good answers come from better-defined questions, and the interesting questions seldom have answers already.
This is an expansion of two themes from my (No) Expectations page.
Teaching
I am not the sole source of knowledge in the world, nor the classroom. I am a guide, a mentor, and a facilitator. I provide a framework for learning, and I suggest robust, yet bespoke, tools to learn. I do not provide (all) the answers. I provide questions and a space to explore them. Sometimes there are known answers, but the interesting questions seldom have those already. I am not the arbiter of truth, but I aim to be a guide to finding it.
My teaching philosophy centers around fostering a dynamic and inclusive learning environment that encourages critical thinking, creativity, and collaboration.
Learning
Learning is available everywhere.
Surround yourself with smarter people.
Be humble, but confident. Be confident, but humble.
Go to a language course for your mother tongue.
Find out new things about the tools you use everyday. Read the readme.
Embrace change. When it is useful. When you are in control, make sure change is useful. Everything new has inertia.
Question everything
Good answers come from better-defined questions. What is asked-for and what is sought are not always the same. I aim to help students, clients, and colleagues find the right questions to ask, and to help them find the answers they seek. I do not provide the answers, but I can help them find them. I do not provide the questions, but I can help them curate them. I do not provide the tools, but I can help them mold them. I do not provide the knowledge, but I can help them seek it. I do not provide the wisdom, but I can help them cultivate it.
Doing is not learning
These are more recent thoughts, sharpened by teaching my Critical AI course.
Producing something is not the same as learning something. We confuse the two constantly, and AI has made the confusion cheap.
“It looks done” is a performance signal, not a learning signal, and the two come apart the moment a tool can finish the artefact for you. The essay used to stand in for the thinking that produced it; that proxy has been severed from what it was pointing at. The uncomfortable part is that the struggle to produce is the learning. Skip the effort and there is often nothing left to remember. A better output today can quietly hide an emptier understanding tomorrow.
You learn through challenges
Difficulty is the mechanism of learning, not the obstacle to it. We learn by working through what is hard; ease is a signal that we already knew enough. Put information-theoretically (because that’s a common lens I use to view the world), you only update where your model was wrong. An easy task carries no prediction error, so there is nothing new to encode.
But (and it is a real but), not all difficulty is good difficulty. Desirable difficulty sits just within reach; difficulty beyond reach is only frustration and wasted effort. And the line between them is not in the task. It moves with the learner. So “how hard should this be?” has no absolute answer: the honest one depends on who is doing it and where they are. Difficulty is a cell, not a number.
This is why I lean on problem-solving and productive struggle: grapple with something hard first, then consolidate. Struggle without consolidation is just failure; consolidation without struggle is just being told. You need both.
Reflection and intentionality
Effort is necessary but not sufficient, so you also have to aim it. This is where intentionality and reflection earn their place, and where they are most often faked.
Reflection after the fact, on its own, has nothing to calibrate against. “What would I have learned?” is easy to answer eloquently and impossible to check. The fix is to reflect against a commitment made beforehand: predict, then check. What matters is the gap between what you expected and what actually happened — not the word-count of the reflection. Well-calibrated people know what they don’t know, which is exactly what tells them when a tool is replacing their learning rather than supporting it.
The good tools (human or artificial) regulate difficulty into that reachable zone and then get out of the way. Scaffold, then fade; support that never fades becomes a crutch. I want tools, and teachers, that think with you, not for you.
None of this is anti-AI. It is anti-shortcut. Used well, AI holds the struggle in the productive range; used carelessly, it deletes the struggle, and the learning with it. The design decides which one you get.
The goal of education is not to produce students who can use AI effectively. It is to produce humans who can think, create, and evaluate — skills that happen to make them excellent users and critics of AI.