Teaching at the African Olympiad Academy

Oct 16, 2025

Introduction

I spent two weeks last month teaching physics at the African Olympiad Academy (AOA), a pan-African, olympiad-focused high school in Rwanda. The school was started just this year by some of the coolest people I know. I first wrote an introduction to this post explaining what the school was all about, because the story and mission of the school are both quite extraordinary. However, I found it a bit cheesy and hard to write, and so ended up scrapping it and will instead redirect you to their website and provide some pictures.

In my two weeks at AOA, I taught 16 two-hour classes (8 unique classes, two times each) to 32 students. I introduced students to the study of physics, and covered motion under constant acceleration, some mathematical) pre-requisites, and the principles behind inertial reference frames.

Below is what I learned from this experience.

Lecturing is surprisingly fun!

This is a lesson I learned about myself, not the Universe. I really enjoy standing in front of the class lecturing. This is slightly surprising because historically, I haven’t exactly been raving to speak in front of large audiences.

I think the main cause for this disparity is that this time, I wasn’t being examined. Most people (myself included) only do public speaking on occasions where there is a clear pressure to perform: oral examinations in school, presentations of work progress to your office team, the list goes on. No wonder public speaking becomes so stressful! Imagine you only ever played football when there was a stupendous amount of money on the line - seems likely that you would end up thinking football is a very stressful and unenjoyable activity. GPT-5 suggests some other areas where this might be true, and they seem reasonable to me.

Other possible reasons: I felt confident in what I was presenting which made it less stressful, and talking to students is more fun than to adults, because they ask more questions.

…but learn when to shut up

That said, about a week in I realized that I had fallen into what I think is a common fallacy for teachers. Namely, spending too much time lecturing to the class, because that is when it feels like knowledge is being conferred from me to the students. This was very much my approach going into teaching - I would prioritize conveying new knowledge during classes, be it through lecturing or demos, and then give problem sets for the students to solve in their free time.

And don’t get me wrong, I’m a huge proponent of having teaching be interactive, and would often ask students to derive parts of the results I was going through in lectures (for instance, doing the algebraic manipulations to derive the suvat equations, after I had set up the premise). But some classes still involved me standing in front of the board for most of the 2-hour long class.

This was a mistake.

The students self-reported that they found these classes to be the most boring ones, and hardest to follow. This particularly hurt the students who were struggling, for whom 2 hours of difficult content would swoosh by very quickly. I tried pretty hard to create an environment that would be very open to students asking all kinds of questions, but alas, this was still a big problem. (And understandably so, asking questions is genuinely difficult! I talk more about this below.) When we moved on to other topics, I could also sense that they hadn’t actually picked up on that much of what I was talking about in the lecture-heavy classes.

So, what’s the alternative? There’s only so much you can do. On the recommendation of my friend and headmaster of the school, Arun, I planned the next few classes to instead be max 1 hour of me lecturing (still interactively) and then 1 hour of students doing individual problem solving with me walking around answering questions. This was much better! Based on feedback, students enjoyed these classes more (even enjoying what I thought would be a rather lame class on reference frames), and knowledge retention seemed better, based on vibes. Some of the students who were struggling now had the confidence to ask me questions individually, and I could also prompt them to ask me questions individually as I walked around the classroom. Finally, this gave me much more signal on how the students were doing - which otherwise is a big challenge (see below).

Coming into my first real teaching experience, my belief in “Make The Teacher Shut Up And Let The Students Talk” as a teaching ideology was already very strong, so I was actually quite surprised to realize that I had fallen into the “Make The Students Shut Up And Let The Teacher Talk” trap this quickly. But I want to note just how counterintuitive the former strategy is when you’re planning a class. For one, it truly feels like you’re imparting knowledge when you blurt it out in front of the entire class (you’re not.) And secondly, it also seems like a much better use of the teacher's resources! If the teacher spends most of the class just idling by their desk, or answering individual students’ questions, they will end up spending a lot of time either doing nothing or explaining things multiple times over.

This was the main reason I initially planned my classes to be so lecture-heavy. After all, the students would solve problems in their free time! I now have more empathy for the teachers of my childhood who would spend long classes speaking in front of class. It really feels like the right thing to do. But it isn’t.

You can just teach things (from first principles)

I really like Socratic teaching. When instructing coaches for the Rwanda Physics Olympiad (my main previous teaching experience), I most often reply to questions with other questions, trying to walk students toward the answer rather than give it to them myself. Whenever I feel the need to give away something, I try only to convey the central intuition or idea behind a proof, and then let whoever I’m instructing write it up themselves. To the extent possible - purity is constraining. When crafting lesson plans at AOA, I tried as much as I could to give students the feeling that they were actual physicists, discovering the hidden truths of the universe for the very first time themselves.

Although I really believe in this style of teaching, I was still a little nervous going in. While Socratic teaching certainly has been tried before, fact of the matter is that basically no schools (ever) operate that way. Maybe there’s some good reason for it? The obvious candidate explanation being that it’s too slow, and there simply isn’t enough time to get through all the material if one is to have students re-derive everything themselves. This is what one of my tutors in Oxford told me, and what most of my friends would say when they objected to my proposed lesson plans.

So, how did it all go?

Well, I actually don’t have that much data, on account of many of my classes still being fairly un-Socratic (shame on me). However, I did make one awfully Socratic class that I was very proud of. The night before, I vibe-coded a web app, which contained a simulation of a ball being thrown up in the air on a mystery planet. Students using the website could change three input variables: initial height, initial velocity, and mass. They could then click “Run” and watch the ball be thrown up into the air and come back down. The question I then posed to the class:

“I want you to tell me how high a ball of mass 25 kg will go if I start it from a height of 25 m, with initial velocity 50 m/s. You have 1 hour to run any experiments you want.”

...

“But sir, we don’t have access to those parameter values in the simulation. For one, height only goes up to 15 m. And we can only input velocities up to 10 m/s.”

*smiles smugly*

Of course, some students tried to apply expressions that they had already learned from studying physics previously. But alas, this mystery planet had a gravity not quite like Earth’s, and so that wouldn’t get them the right answer!

I like this setting because I think a lot of fundamental things about physics are baked into it:

1. Prediction: Physics (and science, more broadly) is about making predictions about what will happen in the world. (Once you have this knowledge, engineering the world to how you want it to be just becomes a search problem over your input space).
2. Compression: To make predictions you can’t just have a lookup table of inputs and outputs. This would be extremely inefficient, and you wouldn’t be able to say anything about cases you’ve never been able to see before. Thus, as a physicist, you seek to find some underlying structure, some governing law, behind your observations. In modern-day ML speak (and old-school information theory speak), you’re compressing your data, if you will.
3. Experiments: How do you compress? Well, you run experiments in the regime you can observe, from which you hope to learn relationships between different variables. For the students, this consisted of varying the input variables in the simulation (say, heigh) and observing how this changed the final height reached by the ball. And here is an important methodological point: Most students did not find the right “trick” of only varying one variable at a time. I had forgotten (or never realized) how unintuitive this idea is!
4. Extrapolation: Then - and here a leap of faith - we extrapolate this relationship, assuming it to hold true even in the regime we haven’t observed. For the students, this was the input values they didn’t have access to. Now, why do we expect this extrapolation to be sensible? Just because final height is proportional to the square of the initial velocity between velocities of 0 and 5 m/s, why should I think that’s true outside of that range? After all, there is no evidence for it! (I was very happy that at least one student asked this). The answer to this has many layers:

   

   1. Empirically, we have observed that usually when we make these simple extrapolations in physics, they hold. This isn’t just true of individual subfields like mechanics, electromagnetism, but everywhere! These observations give us some confidence in thinking that simple extrapolations (almost) always work.
   2. Theoretically, we have reason to believe that we are likelier to be in a simple universe. Hence, we should always guess the simplest thing, which is for laws to be the same everywhere.
   3. Spiritually, wouldn’t it seem strange for the Universe to care about if velocity is greater than or equal to some arbitrary value, and put different laws in place based on that? I mean, that just seems wrong.

Empirically, the students at large enjoyed this class, and another similar one that I gave them the week after. I was quite worried that students would mostly end up just feeling very confused and not really get anywhere at all, learning neither any principles of physics nor any practical takeaways. And while there was some of that, my impression (in particular from the students feedback) was that this experiment was successful, and that the students enjoyed it! I still think it’s a little unclear how effective this was a way of teaching the principles listed above - which I hoped the students would take with them, at least to some extent, from this experiment.

One negative piece of evidence was that the second time that we tried this (the week after) most students still struggled to apply the (relatively) simple method of isolating one variable to vary and controlling the other to, and still seemed quite confused about how to do extrapolations, even after I outlined the method at the start of class and encouraged them to use it. It seems plausible to me that the mean squared error of the final predictions would have been better if I had simply instructed them how to “correctly” do experiments in the first week, and let them try it themselves the second week. But on the whole, I take this as a (slight) win for Socrates.

Making students comfortable to ask questions takes time and effort

I really wanted to create a culture where students would feel welcome to ask any and every question they had, particularly when they didn’t understand things. To this end, I came up with a game that I would play with the students at the start of every class, where I forced them all to come up with a question (on any topic!) and then I would pick a few of them and ask it to me. My hope was that this would a) signal that I really endorse asking questions, even once that seem silly and b) get the students to do active inference on actually asking questions at the start of each class. It was a nice idea, but in practice it didn’t seem to transfer very well into asking questions in class, and students still at large did not feel comfortable asking questions in class when they were confused. So, this requires work. My thoughts on the right approaches:

1. Be explicit in expressing your desire for questions. At regular intervals in class time, stop and say: “You should probably ask me questions on this right now.” (shoutout to David Quarel, from whom I first heard this).
2. Express appreciation for questions when you receive them, and always try your best to answer them when appropriate!
3. Create spaces for students to ask questions individually. For instance, by setting aside class time for individual problem solving where you walk around and answer questions (prompt students to ask if necessary). This will help build their confidence.
4. Try to identify any other underlying reasons for why the students aren’t asking questions, and address them. Is the classroom dominated by a few students asking very advanced questions? Are classes overly goal-focused, making students feel like detours are not welcome?
5. (maybe) Do some activity that you hope will transfer - this is pretty cheap and might work.

As a teacher, you are really constrained for feedback signals

Getting feedback from most humans is difficult. Getting it from students is even harder, for a few reasons. For one, most students are focused on how they can improve, not how their teacher can improve (constructively, at least). Secondly, even if students have ideas, they’re often afraid to say so (even in anonymous feedback forms!) because they are determined to stay on the teacher’s good side - this is surprisingly common even in standardized exam systems, where the teacher isn’t the one assigning the final mark students get! Here are my best tips:

1. Make sure you get a chance to help students individually in each class. This gave way more signal on how much they were understanding than my interpretation of their facial expressions during lectures.
2. If you ask for a feedback form (which I did at the end of my two weeks), ask questions like: “If you had to change something, what would you change?”. I think this worked much better than “Is there anything you would change?” would have.
3. Make sure you update on feedback, that will make students more likely to give it. When I announced my feedback form, I really emphasized that I was intending to change my classes based on the feedback they gave - many students seemed to react (positively) to this, but I’m unsure if this had any effect.
4. Try to meet with students individually every so often (perhaps once a term or once a year), and spend 5-10 minutes discussing their thoughts and elicit their feedback in a more low-stakes and personal setting. (I didn’t try this, but a friend recommended it and it seems very reasonable to me.)

Don’t mark any homework yourself

Marking a pretty light problem set completed by all 30 students took me about 6 hours, and I ended up writing basically identical feedback on most questions, since most mistakes were common. For reference, planning a class or making a problem set usually took me about 2-3 hours, and I think the marginal value of my time was much higher for that compared to marking, where I added relatively little value.

I would recommend anyone who does teaching to not do any marking themselves and instead just let students self-mark, or mark each others’ papers in class, and then go through all the questions together. While it’s nice to be able to give individualized feedback, marking assessments individually for each student doesn’t seem like a particularly effective way to do this. It’s asynchronous which means you can’t interact with the student, and so many mistakes are shared across students for most assessments. So, I would recommend prioritizing that time toward open office hours or small-group tutoring with students instead. With 6 hours of your time, you could split a class of 30 into 6 groups of 5 and have 1-hour long tutorials with each of them instead, which seems way more valuable.

Answering all kinds of questions is appreciated

One of the most common pieces of positive feedback I got from students was that they appreciated that I tried to answer every single one of their questions, even ones that were on topics far beyond what we were studying (“Why doesn’t our galaxy collapse under its own gravity?”)

Doing the same class twice can get tiring

To combat this, try to find a way to spice up the classes (for yourself, not just the students)! My main remedy for this was to not have a script and allow myself to freely make comments and jokes. This helped somewhat.

Postlude

I really enjoyed my two weeks at AOA. Since leaving, I think I have done a poor job of conveying to my friends and family just what an amazing place I think it is. And I’m afraid this set of reflections on teachings has also at large evaded that topic - partly because there isn’t much to say about it. Nevertheless, these reflections would certainly be incomplete if I didn’t offer at least a few words on the topic.

I thought (and still do) that my high school, UWC, was the best place and collection of people on Earth. My high school had it. It had the kind of culture where the whole school would cheer on the shiest student making their first performance at the annual talent show. It had the collective enthusiasm which made us all get up and sing happy birthday in the school canteen whenever someone had a birthday. It had the strength of community which to this day makes adults who graduated decades ago run across airport terminals just because they saw someone with a UWC hoodie, even if they attended years apart.

When I was at AOA last month, I felt - and remembered - feelings that I hadn’t since I left my high school. I don’t know how the founding team did it. But AOA has it. The day I left, there was the first school-wide gathering, an Indaba, to discuss school-wide events and bring up any other important information. At one point, the students were asked if they had anything to share. For 20 minutes, students got up one after another - unprompted by any adults - and express their gratitude toward the staff, the teachers, and most of all, each other. They had also made me a beautiful cardboard book with written notes from all of them, and one student had written a poem - the kind of gift which gives me joy even as I think back on it now. I thought there was only one place in this world where students did that, and that I had already spent my late teenage years there. Turns out there are two - and I feel lucky to have seen them both.

I am grateful to Vincent Cheng, Arjun Khandelwal, Rohan Selva-Radov, Yonatan Gideoni, & Nick Marsh for helpful comments.