You might have seen in the news that ChatGPT recently solved a math problem that has confounded mathematicians for centuries.

You might rightly wonder: in a world where AI performs at the level of the world's top mathematicians, why are we even teaching math in schools?

Or writing? Or science? Or any of the academic subjects where AI now performs at the level of a PhD, or better?

At Lumos, we spend a lot of time thinking about this. In a world of AI, what is school for? What kind of education will give our students the greatest chance of success in their adult lives?

Children's time is precious. We don't want them doing things simply because that's how school has always worked.

Keep doors open. But don't stop there.

One answer is practical: academic foundations and qualifications still open doors to secondary schools, apprenticeships, universities and other opportunities.

Those options matter to us. We want Lumos students to be able to pursue ambitious academic paths, including leading universities.

But preparing for the next exam and getting good grades cannot be the whole reason for spending thousands of hours studying.

Illustration: Traditional school teaches kids to do what they’re told. AI already does that. High-agency kids know what to do when nobody tells them.

We're not just teaching answers. We're developing thinkers.

At Lumos, we think the primary purpose of academics is to train the kind of mind that will lead AI, not be led by it.

We've all seen AI slop. The kind of mediocre output AI produces when it's directed by someone who's happy to accept the first answer they're given. Who doesn't write precise, thoughtful prompts or hold a high bar for quality writing and rigorous arguments. Who settles for generic ideas instead of original ones. Who accepts what's given instead of asking: does that really make sense?

A mediocre mind produces mediocre results, even with powerful AI. A thoughtful, critical, well-trained mind produces high-quality results. The first outsources its thinking to AI. The second turbocharges its thinking with AI.

Building strong minds requires doing hard things.

How do we train children to develop the second kind of mind?

By making them go through the hard, deliberate, iterative (but ultimately rewarding) work of doing academics without the assistance of technology.

Math develops logical thinking, problem-solving and persistence. The child who works through challenging, open-ended math problems learns persistence, problem-solving and comfort with mistakes. The child with strong number sense and the ability to do quick mental calculations is less likely to fall for misleading information or be manipulated by data. The child who understands probability will make better life decisions. Math, taught well, develops logical and systematic thinking, skills that are valuable no matter where AI takes us.

Writing develops clear thinking, precise communication and a unique voice. The child who learns to write well independently will recognize what good writing looks like. By doing the hard work of developing their own unique voice, they can later leverage AI without losing it. The alternative is generic AI writing, where everyone sounds the same and no one sounds authentic. Learning to write well also leads directly to speaking well. It trains children to organize their thoughts, structure arguments, use evidence, persuade and choose their words carefully.

Illustration: Both routes can end in the same essay. Only one changes the child, because the struggle is where the learning happens.

Science and history develop critical thinking and a deeper understanding of the world. The child who learns to think like a scientist learns to seek verifiable truth, gather evidence, change their mind when presented with new information and think critically. The child who studies history is better equipped to understand the dynamics of the world around them. And so on.

The greater risk isn't that children won't know how to use AI.

We think we're at far greater risk of raising a generation of kids who can't think for themselves and turn to technology as soon as something gets challenging than a generation who can't leverage AI effectively.

Enough to pass a test. But enough to build a mind?

If the purpose of academics is to develop strong thinkers, do we need to change how we teach them?

Absolutely!

When exam results are the main measure of success, relatively shallow learning can look like achievement. When the goal is independent thinking, that's no longer enough.

At Lumos, math goes much deeper than teaching kids to reproduce the teacher's method. Our students tackle math puzzles and logic problems, some of which may take weeks to crack. We dedicate time to developing mental math so students can quickly judge whether numbers make sense.

When introducing new concepts, we don't take a "show, mimic" approach, where the teacher demonstrates and students replicate. Instead, we help students understand why something works. This leads not only to deeper, more lasting conceptual understanding, but also develops persistence, problem-solving and logical thinking.

Writing at Lumos is an exercise in iteration. Students never stop at one draft. They're asked to rewrite sentences, choose better words, find stronger evidence, cut entire paragraphs in pursuit of brevity and anticipate counterarguments. This helps them become not just stronger writers, but clearer thinkers and more persuasive, articulate communicators. The ability to give precise instructions and iterate until you achieve a quality result is also foundational to using AI effectively.

Across all subjects, we ask ourselves: what's the deeper purpose of learning this? What enduring habits of mind and thinking skills will students develop? How will it help them become more discerning, critical thinkers?

Then we design our lessons with those overarching goals in mind.

But it can't just be about academics, can it?

Delivering deep, engaging academics is something we care a lot about. But it's far from our only priority.

We believe it's at least as important, perhaps more so, to help students develop excellent interpersonal and collaboration skills, a strong sense of self, robust executive functioning skills, and the ability to persist and adapt in the face of challenges.

This is one of the primary reasons we have such a low student-to-teacher ratio and focus on one-on-one teaching. Numerous studies have demonstrated the effectiveness and efficiency of personalized tutoring. By taking this approach, we free up meaningful time in our weekly schedule for other pursuits:

• Passion Projects: Students pursue real projects, set goals and timelines, adapt when inevitable roadblocks arise and present their work to real audiences.

• Agora: A daily practical workshop to develop critical thinking, emotional intelligence, teamwork and technical skills.

• Journaling: The bookends of each day, when students reflect on their learning and habits.

• One-on-one coaching: Weekly sessions where students reflect on their progress and development with a dedicated coach.

• Expeditions: Three week-long experiences each year designed to take students out of their comfort zones, build independence and develop resilience.

Not competing against AI. Learning to thrive with it.

AI doesn't make education obsolete. It asks us to be clearer about what education is for.

Our goal is to develop curious, knowledgeable, persistent young people who can use powerful tools without becoming dependent on them. Children who can question an answer, improve it, reject it, and know when to put the technology aside and think for themselves.

That's the kind of education we believe is worth their time.