and the AI Revolution
In 2026, we're arguing about whether AI is ruining mathematics.
Perhaps we're asking the wrong questions.
๐ On the Starship Enterprise, You Still Have to Learn Calculus
In a 1988 episode of Star Trek: The Next Generation titled When the Bough Breaks, a young boy named Harry announces that he hates calculus and doesn't want to study it anymore.
His father reminds him that everyone needs some understanding of calculus. By the end of the episode, Harry explains that he wants to become an artist instead.
His father's response?
"You can be anything you want, Harry. Anything. But you still have to take calculus."
I love that scene.
Here we are, imagining life aboard a 24th-century starship, surrounded by extraordinary computers, intelligent machines, voice-controlled systems, simulations, and technology capable of calculations that would make today's supercomputers look primitive.
And children still need to learn calculus!
But here's what I find particularly fascinating.
The crew of the Enterprise doesn't spend its time sitting around solving pages of derivative problems or calculating integrals by hand. When confronted with a challenge, they use their technology. They investigate, model, simulate, calculate, experiment, collaborate, and make decisions.
They use mathematics to accomplish something.
When Geordi La Forge encounters an engineering problem, he doesn't reach for a worksheet. He works with the ship's computer, examines data, tests possibilities, and adjusts his approach. Data, an android with extraordinary computational abilities, isn't considered a threat to the crew's mathematical competence. He's an invaluable collaborator.
The technology expands what people can accomplish.
And the episode itself offers an interesting warning: On the planet Aldea, a civilization has become so dependent on its automated technology that its people have lost much of their understanding of how their own systems work.
There's a lesson in that, too.
Now fast-forward from 1988 to November 2026.
We have artificial intelligence capable of solving increasingly sophisticated mathematical problems, and much of mathematics education seems to be having an identity crisis.
What a mess.
๐ณ Two Stories from Caltech That Say Everything About Our Dilemma
Two recent stories about Caltech capture the contradictions beautifully.
A Caltech Professor No Longer Grades Homework for Accuracy
Students can use AI to solve traditional homework problems, raising questions about what homework grades actually measure.
Read the Business Insider story →On November 16, 2026, Business Insider published Truman Dickerson's article about attending a mathematics class at Caltech. The professor, Yi Ni, reportedly no longer grades homework for mathematical accuracy because students can use AI to solve the assigned problems. Instead, he grades homework for completion (Dickerson, 2026).
The article describes a traditional classroom: a professor filling blackboards with mathematics while students take notes.
The technology has changed dramatically.
The classroom? Not nearly as much.
The professor recognizes that AI can solve much of the homework, but the educational response described in the article is to stop grading its accuracy. Students still face traditional examinations, where understanding is expected to show up without AI assistance.
I don't want to dismiss the professor's concerns. They are legitimate. Students need to develop mathematical understanding, and outsourcing every intellectual challenge to an AI system is hardly a recipe for learning.
But I can't help wondering:
If a computer can complete the homework, is the real problem the computer—or the homework?
Now consider the second story.
Caltech and AIM Build an AI Platform for Mathematicians
An initiative to develop AI tools that support mathematical research, reasoning, conjectures, and collaboration.
Read Caltech's announcement →On October 7, 2026, Caltech announced a collaboration with the American Institute of Mathematics to develop an AI platform specifically designed to help mathematicians conduct mathematical research.
The platform is intended to help researchers connect ideas, formulate conjectures, explore strategies, organize knowledge, and pursue mathematical reasoning collaboratively (California Institute of Technology, 2026).
Think about that.
At the same institution, we see two very different responses to the same technological revolution.
To be clear, these stories describe different mathematical activities with different educational needs. Researchers possess knowledge that beginning students are still acquiring. And Caltech's research initiative is a platform under development, not proof that AI has solved the challenges of mathematical discovery.
Still, the contrast is remarkable.
One part of the institution is trying to figure out what to do when machines can complete assignments.
Another is building machines to help people do mathematics that hasn't yet been done.
Perhaps the problem isn't that AI is changing mathematics. Perhaps the problem is that mathematics education hasn't changed enough.
Are We Teaching Students to Do Mathematics—or Simply to Find Answers?
This is the question I keep coming back to.
For generations, a substantial portion of school mathematics has followed a familiar pattern:
The teacher explains a procedure. Students practice that procedure. Students receive a collection of similar problems. Eventually, they take a test demonstrating that they can reproduce the procedure independently.
Success often means getting the right answer using the expected method.
There's nothing inherently wrong with developing procedural fluency. Students need foundational knowledge. Calculating, manipulating expressions, understanding functions, and recognizing mathematical relationships are important.
But we have allowed procedural performance to become far too closely associated with mathematical competence.
And AI is exposing that weakness.
If our primary educational objective is to train students to produce answers to existing mathematical problems, we are now competing with machines designed to produce those answers more quickly.
That is not a competition education can win by simply writing longer tests, assigning more homework, or banning technology.
More importantly, it is not a competition we should want to win.
Consider what mathematicians, scientists, engineers, architects, designers, and entrepreneurs actually do.
They encounter situations where the problem may not be clearly defined. The necessary information may not be available. Multiple approaches may be possible. Assumptions need to be made and tested. Solutions must work within real constraints.
Sometimes, the most difficult part isn't solving the mathematical problem.
It's figuring out what mathematics is needed in the first place.
I worry that many of our students rarely experience that part of the discipline.
And that brings me to an uncomfortable thought.
๐ช What If Some Math Teachers Have Never Really Experienced Doing Mathematics This Way?
This question may make some educators uncomfortable.
It should.
How many mathematics teachers have spent most of their academic and professional lives working on problems for which someone already knew the answer?
They learned mathematics by solving assigned problems, passed examinations by reproducing established techniques, and entered classrooms where they were expected to teach the same techniques.
They may be exceptionally skilled at algebraic manipulation, differentiation, integration, and explaining mathematical procedures.
But have they had opportunities to use mathematics to create something original?
This isn't an accusation that mathematics teachers don't know mathematics. It is a challenge to our definition of mathematical expertise—and to the opportunities we give teachers to develop it.
Knowing how to solve a differential equation is valuable.
Knowing when a differential equation might help solve a meaningful problem is another kind of expertise.
Knowing how to use computational technology to explore that equation, interpret the results, and make a decision is yet another.
We need teachers who are not only skilled at explaining mathematics but also comfortable creating situations in which mathematics becomes necessary.
That is a different kind of teaching.
And it may require us to become learners again.
๐ ️ Stop Assigning Problems. Start Giving Students Challenges.
Imagine walking into a mathematics classroom and hearing the teacher say:
Your challenge is to design and build something that works. You can use mathematics, AI, simulation software, spreadsheets, 3D printers, coding tools, and whatever other resources you can justify. But you must explain the mathematics behind your decisions, demonstrate that your design works, and convince us that your solution makes sense.
Now we're getting somewhere.
What if the challenge were to design a solar-powered charging station for the school? Students would need to investigate electrical consumption, solar exposure, panel orientation, efficiency, cost, and energy storage.
Suddenly, algebra, trigonometry, functions, and optimization become useful.
Not because a textbook says they're important.
Because the students need them to accomplish their goal.
Or imagine asking students to design a small roller coaster, build a working scale model, create an interactive computer simulation, or engineer a rainwater collection system.
The mathematics isn't presented first as an abstract procedure to memorize.
It emerges as a tool students need.
Four Challenges That Could Transform a Math Classroom
Students design a physical or virtual coaster, using functions, derivatives, slope, curvature, energy, and constraints to improve their designs.
Students use data, trigonometry, optimization, estimation, and mathematical modeling to develop a feasible proposal.
Teams design, simulate, build, test, and refine structures using geometry, measurement, ratios, optimization, and statistical comparisons.
Challenge students to identify a real problem and design a useful object to solve it: a custom phone stand, an assistive device, a replacement part, or an organizational tool. Students must apply geometry, measurement, proportional reasoning, surface area, volume, and optimization to develop their designs. They create digital models, produce physical prototypes, test their functionality, and revise their designs based on the results.
The mathematics matters because the finished object actually has to work!
The point isn't that every classroom needs expensive equipment or a semester-long engineering project.
It's that students need experiences where mathematics serves a purpose beyond producing an answer on paper.
A mathematical model, an interactive visualization, a functioning prototype, a useful recommendation, or an evidence-based decision can all become meaningful products of mathematical thinking.
This isn't merely an attractive educational philosophy.
The National Council of Teachers of Mathematics identifies modeling as a process that starts with real situations, requires assumptions and mathematical representations, and involves evaluating and refining solutions (NCTM, 2020).
Research by Turner et al. (2024) documents how teachers created authentic modeling experiences around cafeteria waste, community resources, and a school art project.
And Hofmann et al. (2026) describe computer-aided mathematical modeling workshops involving thousands of students, offering another example of how mathematical activity can connect with contemporary technological practice.
These studies don't establish that projects should replace all explicit instruction. They do, however, illustrate that authentic mathematical work is possible—and that we have models worth learning from.
๐ค AI Isn't the Enemy. A Lack of Purpose Is.
Here's an idea that may sound radical.
What if students were encouraged to use AI when working on meaningful mathematical challenges?
Not to copy answers.
Not to avoid thinking.
But to help them investigate possibilities, generate code, test models, analyze data, identify errors, and explore mathematical ideas beyond what they could accomplish alone.
That's what technology should do.
A student designing a solar charging station might ask AI to help create a simulation of daily energy production. But then the student must understand the variables, verify the calculations, identify questionable assumptions, and compare predicted performance with real observations.
If the simulation suggests an impossible result, who catches it?
If the design goes over budget, who decides what to change?
If the mathematical model doesn't fit the observed data, who revises it?
AI can assist with the computation. It cannot be allowed to take responsibility for the student's thinking.
The educational challenge is to make that thinking observable.
Students can document revisions, defend assumptions, explain their models, respond to unfamiliar questions, and demonstrate that their work functions as intended.
And yes, they should sometimes work without AI. Just as musicians practice scales and athletes develop basic techniques, students need opportunities to build foundational mathematical fluency independently.
But those opportunities should support richer mathematical work, not define its limits.
The Real Challenge: Changing What We Assess
This may be the most difficult part.
We can introduce AI, modeling software, and creative projects into mathematics classrooms tomorrow.
But if success is still determined almost entirely by performance on a timed written examination, what message are we sending?
We are telling students that the project is interesting, but the test is what really matters.
And teachers, understandably, will continue preparing students for the test.
This is not simply a classroom problem. It is a systemic one involving curriculum requirements, university admissions, standardized assessments, schedules, and deeply held assumptions about what mathematical competence looks like.
Still, we can begin changing what counts as evidence of learning.
Imagine assessing students on their ability to define a problem, select appropriate mathematics, construct and test a model, evaluate the limitations of their results, and communicate their reasoning.
Written examinations can remain part of that assessment. But so can demonstrations, oral defenses, design portfolios, prototypes, and explanations of failed attempts.
A student who builds a mathematical simulation and can explain why it works—or why it initially failed—has demonstrated something valuable.
We should recognize that.
A Different Set of Questions for Math Assessment
| Instead of Only Asking… | Also Ask… |
|---|---|
| Did you get the correct answer? | Does your solution work under realistic conditions? |
| Did you use the required procedure? | Why did you select this mathematical approach? |
| Can you solve it independently? | Can you use tools responsibly and verify their results? |
| Can you complete the test? | Can you defend, improve, and communicate what you created? |
๐ฌ Of Course Teachers Are Resistant. This Is Hard.
I understand why changing mathematics education can feel threatening.
Teaching through authentic challenges is messy.
The teacher may not know the answer in advance.
Students might pursue unexpected approaches.
Technology might malfunction.
A project might fail.
The mathematics needed might extend beyond the originally intended curriculum.
And perhaps most uncomfortable of all, a student might ask a question the teacher cannot immediately answer.
But isn't that exactly what authentic mathematical work looks like?
The teacher's role shifts from being the person who knows all the answers to being someone who can guide inquiry, recognize mathematical opportunities, ask productive questions, and help students make sense of what they discover.
That's demanding work.
It requires significant professional development, planning time, curricular flexibility, accessible technology, and institutional support. Simply telling teachers to be creative while leaving everything else unchanged is neither fair nor realistic.
But the alternative—continuing to prepare students for a mathematical world that no longer exists—is becoming harder to defend.
Perhaps teachers should experience this kind of learning themselves.
Imagine mathematics professional development where teachers are challenged to design a working irrigation system, build a simulation, optimize a physical structure, or investigate a community problem.
Let them work with AI.
Let them encounter uncertainty.
Let them experience the frustration of a model that doesn't work—and the satisfaction of finally figuring out why.
Back to the Enterprise: The Future Isn't About Doing Everything by Hand
I keep returning to that young boy on the Enterprise who wanted to be an artist instead of studying calculus.
What if his teacher had asked him to design a new sculpture for the ship's observation lounge?
Perhaps the sculpture would involve curved surfaces, changing dimensions, or movement.
He could use the ship's computer to model his design. Calculus could help him understand the curves, optimize material use, or describe how the sculpture changes over time.
He would still need to understand calculus.
But now he would have a reason to care.
His artistic ambitions and mathematical education wouldn't need to compete with each other.
They could strengthen each other.
That's the future of mathematics education I would like to see.
Not a future where everyone sits at a computer and asks AI for answers.
Not a future where students never learn to calculate, reason, or prove.
And certainly not one where technology replaces genuine intellectual engagement.
I want classrooms where students use mathematics to create things, investigate problems, pursue ideas, and accomplish goals that matter to them.
Where technology expands their possibilities.
Where getting the right answer is important, but knowing what question to ask is often even more important.
Where students encounter mathematics as something people actually do, not merely something they complete.
๐ It's Time to Boldly Go Where Math Education Has Been Reluctant to Go
The two Caltech stories offer a glimpse of a crossroads.
On one side, we have an educational system struggling because AI can solve the problems we assign.
On the other, we have mathematicians exploring ways to use AI to pursue problems that may not yet have solutions.
We need to bring those worlds closer together.
I am not suggesting that eighth graders should conduct advanced mathematical research or that college calculus courses should become engineering workshops.
I am suggesting that every student deserves opportunities to use mathematics to investigate, design, create, and make sense of something that doesn't come with an answer key.
Start small.
Replace one routine assignment with a meaningful challenge. Give students a goal and some constraints. Allow appropriate technology. Require mathematical explanations. Let students revise their approaches.
And see what happens.
We may discover that students are capable of far more mathematical thinking than our traditional assignments have allowed them to demonstrate.
We may also discover that we, as educators, have much to learn.
There is uncertainty in all of this. There will be resistance. There will be mistakes.
But those aren't reasons to preserve an increasingly outdated instructional model.
They are reasons to start experimenting.
After all, if the crew of the Enterprise had responded to every unfamiliar challenge by insisting on doing things exactly as they had always been done, they wouldn't have made it very far.
And perhaps that's the real final frontier.
๐ฌ Let's Continue the Conversation
When was the last time your students used mathematics to create something where neither they nor you knew the answer in advance?
And what would it take to make that experience a regular part of mathematics education?
Share your thoughts and classroom experiences in the comments.
๐ References
California Institute of Technology. (2026, October 7). Caltech and AIM build AI platform with mathematicians for mathematicians.
Read source
Dickerson, T. (2026, November 16). I went to a math class at Caltech. The professor no longer grades homework because everyone's using AI. Business Insider.
Read source
Hofmann, S., Bata, K., Schรถnbrodt, S., Kindler, S., Frank, M., Mรผller, C., Eckert, J., Hรถrter, J., Bรผsing, C., & Stamm, B. (2026). Mathematical modeling in action: Design principles for mathematical modeling days. Proceedings in Applied Mathematics and Mechanics, 26(2), Article e70157.
https://doi.org/10.1002/pamm.70157
National Council of Teachers of Mathematics. (2020). Standards for the preparation of secondary mathematics teachers.
Read standards (PDF)
Shearer, H. L. (Writer), & Manners, K. (Director). (1988, February 15). When the bough breaks (Season 1, Episode 17) [TV series episode]. In Star Trek: The Next Generation. Paramount Television.
Turner, E. E., Bennett, A. B., Granillo, M., Ponnuru, N., Roth McDuffie, A., Foote, M. Q., Aguirre, J. M., & McVicar, E. (2024). Authenticity of elementary teacher designed and implemented mathematical modeling tasks. Mathematical Thinking and Learning, 26(1), 47–70.
https://doi.org/10.1080/10986065.2022.2028225
