AI Olympiad: How Do You Stack Up? A Free Assessment and Report

Students preparing for the AI Olympiad have been taking a starting assessment designed by our Mentors in Residence. It covers Python, mathematics, AI coding, and AI concepts. The results say a lot about where students begin, and one lesson stands out: no two students start in the same place. Some are surprised by where their gaps turn out to be. Where would yours land? The assessment isn't public, but your student can take it for free, with a personal report delivered live and 1:1.

An assessment built by people who know AI

Our Mentors in Residence chose every question. They picked the ideas that matter most for the AI Olympiad, and the mistakes they see students make most often in AI. The assessment isn't published or available online, so the only way to take it is through us.

That's the advantage. A practice test found online tells a student which answers they got wrong. This one shows where their understanding breaks down, in the areas the Olympiad actually tests.

What the assessment covers

The assessment works through four areas, each moving from a warm-up to harder, multi-step problems, for example:

- Python: writing short functions in a notebook, from counting and filtering lists to grouping records and computing averages

- Mathematics and Linear algebra: vector arithmetic, dot products, distances, matrix-vector multiplication, and writing one vector as a combination of others, outer products, rank, and how keeping only the most important "pattern pieces" compresses an image

- AI Concepts: how a language model splits text into tokens, handles rare words, and pads examples of different lengths

The problems adapt as a student goes. Someone who handles the early questions well moves on to harder ones, so no two students see exactly the same assessment.

What we found

1. The range is wide

Results varied widely, from students still building foundations to students already working comfortably at Olympiad level. Some moved quickly; others took their time. Every student showed real strengths, and every student had something to work on.

That spread is the whole point of a starting assessment. A student who wants to target the Olympiad needs to precisely identify the areas to focus improvement in.

2. Profiles are jagged, not flat

Overall scores hide the most useful part. Students who were strong in one area were often much weaker in another:

- Some students handled linear algebra flawlessly but struggled with the Python problems.

- Others wrote clean, correct code but stalled on matrix decomposition.

- Several did well on the concept questions about modern AI while making errors on basic vector arithmetic.

A single "readiness score" would miss all of this. Preparation works best when it targets the actual gaps.

3. Modern AI concepts came easier than the foundations

Perhaps surprisingly, tokenization (how models like ChatGPT read text) was the strongest area overall. Python was the area where students most often had room to grow.

Our impression is that students absorb AI vocabulary from the world around them, but the foundations underneath, such as writing correct code and multiplying matrices by hand, take deliberate practice. The Olympiad tests both.

4. Some mistakes came up again and again

A handful of misconceptions accounted for many lost points. Just a couple of examples/

- **Matrix-vector multiplication.** The most common error was multiplying entry by entry instead of taking each row's dot product with the vector, which produces a grid of numbers instead of a vector.

- **Values vs indices in Python.** Writing `for i in nums:` and then using `nums[i]`, treating each number as a position in the list.

**Small slips with big consequences.** Using `>` instead of `>=`, starting a counter at 1 instead of 0, or naming a function differently from what was asked. Each would fail an automated grader.

5. The right answer isn't always the right understanding

Many multiple-choice answers came with no explanation. Among those that did explain, we sometimes saw a correct choice reached by the wrong method, such as computing a dot product as a distance and happening to land on the right option.

On a multiple-choice exam, that still scores a point. But the next problem built on the same idea will expose the gap. That's why our assessment asks students to explain their thinking, and why we look at the reasoning, not just the answer.

6. "I don't know" is useful information

Some students simply wrote "not sure" or "I haven't learned this yet." That's not a failure. It's exactly what a starting assessment is for: finding out what hasn't been taught yet, so it can be, rather than guessing.

Olympiads are relative

An Olympiad isn't graded against a fixed bar. Students advance and earn recognition by how they place compared with everyone else taking the exam. A strong score in a strong year can count for less than the same score in a quieter one.

The AI Olympiad is still young, and early on, fewer students were preparing seriously. That's changing quickly. Each year, more students arrive with more preparation, and the bar moves up with them. Knowing where your student stands on their own is useful. Knowing where they stand relative to other well-prepared students is what matters on exam day.

What the report tells you

A score on its own doesn't mean much. What matters is what it means for *your* student, and how it compares.

That's where our Mentors in Residence come in. They've worked with many students preparing for AI competitions, so they can read a result both on its own terms and relative to other students at a similar stage: whether a gap is typical at this stage or worth attention, whether a strength is ordinary or unusual, and which areas will make the biggest difference with the time available. That judgment is built into each student's report and the conversation that goes with it.

What you get, free

- The assessment itself, designed by our Mentors in Residence and not available publicly

- A personal report on your student's strengths and gaps, area by area, read in the context of our mentors' experience with many other students

- A 1:1 conversation with a member of our team to walk through the results and what to work on first

There's no cost for the assessment, the report, or the conversation.

Why it helps to know now

Every student has gaps. The only question is when you find them, and whether you find them before the students you're competing with find theirs.

Finding them early is almost always better. A gap found in the fall leaves months to close it, calmly and in the right order. A gap found close to the exam leaves little time, and one found in the exam room leaves none.

Knowing also changes how a student spends their time. Instead of reviewing what they already know, they can work on what actually moves their score. It costs little to find out, and not knowing costs far more.

What this means for preparation

- Start from where you actually are. A plan built on a real assessment beats a generic study schedule.

- Don't skip the foundations. Familiarity with AI ideas is a good start, but Olympiad problems also require clean code and fluent linear algebra.

- Always write down your reasoning. It's the fastest way to catch a lucky guess before an exam does.

- Practice the small things. Off-by-one errors, sign errors, and returning instead of printing are fixable with deliberate practice.

Want to know where your student stands?

Book a free consultation. Your student takes our Mentors in Residence's assessment, and you get a personal report, delivered 1:1, on their strengths, their gaps, and what to work on first. All of it is free. The earlier you know, the more time there is to act on it.

Book a free consultation here.