What Can a Student Learn by Teaching a Computer?

From the First Five to STEAM, systems thinking, and AI readiness

When I got my first computer for Christmas in 1983, I was not thinking about artificial intelligence, computational thinking, STEAM education, or AI readiness.

I was trying to figure out what I could make a computer do.

I taught it to say “Hello World.” I programmed it to play “Jingle Bells.” I used BASIC to draw the Columbia Space Shuttle, partly because I dreamed of becoming an astronaut. I taught it to do simple mathematics. Eventually, I programmed a game of Roulette.

Those experiences became the foundation for something I would articulate decades later as The First Five Things to Teach a Computer:

  1. Communication
  2. Music
  3. Art
  4. Mathematics
  5. Gaming

The First Five is my original framework. It grew from my own experiences learning by teaching my first computer. I first published the framework in 2024 in The First 5 Things to Teach a Computer, where I described students taking on the role of teachers and computers becoming their “digital pupils.” 1

That relationship is the idea I have continued to develop.

What can a student learn by teaching a computer?

There is a long tradition of asking how computers can help teach students. The First Five turn that relationship around. What happens when students must explain something clearly enough for a computer to act upon it, observe what happens, recognize when something is wrong, and try again?

The computer may be the pupil in that exercise, but the human is the one learning.

The claim is not that every interaction with a computer produces learning, or that a computer can replace a teacher. The learning depends on the work the student must do: define a goal, make assumptions visible, test an output, explain a judgment, and revise an approach. The computer supplies a responsive medium and a source of feedback. The student supplies the purpose, context, and responsibility.

Technology as a Tool

For years, one of the objectives on my professional goal sheet was simple:

Accelerate Learning with Technology as a Tool.

I have come to appreciate the importance of the final three words: as a tool. Technology was never supposed to be the destination. Learning was.

That distinction becomes more important as technology becomes more capable. My first computer required BASIC commands. Later came networks, the Web, databases, cloud computing, data science, and machine learning. Today, students can interact with generative AI systems capable of producing language, images, software, analysis, and ideas.

Tomorrow’s tools will be different again. If education is organized primarily around learning today’s technology, some of what we teach will age along with that technology. But if students learn to use technology to investigate, communicate, calculate, create, design, model, experiment, evaluate, and solve problems, those capabilities can travel with them from one generation of tools to the next.

That is the thread connecting my early programming experiences with how I now think about artificial intelligence.

Learning is the purpose. Technology is the tool.

The First Five

The First Five began with five things I remember teaching my first computer.

1. Communication

I taught the computer to say “Hello World.”

At the simplest level, this was output, but it also introduced a relationship between human intention, instructions, and machine response.

2. Music

I programmed “Jingle Bells” in BASIC.

Music introduced sequence and timing: the instructions had to occur in the right order and produce the right relationships for the computer to create the melody.

3. Art

I used BASIC to draw the Columbia Space Shuttle.

That required coordinates, geometry, spatial reasoning, and the translation of an idea into instructions a computer could execute.

4. Mathematics

I taught the computer to perform simple calculations.

That introduced precision, logic, abstraction, and the relationship between an expression and its result.

5. Gaming

I programmed a game of Roulette.

Games brought many of the previous ideas together: communication, rules, sequence, logic, interaction, and randomness.

When I first published the framework, I described these experiences as building blocks of computational thinking. They encourage students to deconstruct problems, think algorithmically, and approach challenges through a computational lens. 1

But the longer I have worked with the First Five, the more I have come to see another layer. The computer connected all five experiences, yet it was not necessarily what I was learning about. It was something I was learning with.

The First Five were not simply five lessons about computers. They were five different ways of using a computer to explore ideas.

That distinction provides an important bridge between technology as a tool and STEAM.

The First Five Through a STEAM Lens

I did not derive the First Five from STEAM, nor did I describe them as a STEAM framework when I first developed them. The relationship is one I recognize in retrospect: the interdisciplinary character was already present in the experiences themselves.

Communication crossed language and technology. Music combined creativity with patterns, sequence, and timing. Art connected visual expression with coordinates, geometry, and representation. Mathematics introduced logic and abstraction. Gaming combined rules, probability, interaction, strategy, and feedback.

The five were never clean disciplinary silos. STEAM provides a useful contemporary lens for understanding why that matters. Integrative approaches to science, technology, engineering, arts, and mathematics deliberately create connections among disciplines rather than treating each as an isolated body of knowledge.2

That also helps clarify something important about the role of technology. Technology can be something students study, but it can also be something students study with.

A learner might use technology to:

  • observe a scientific phenomenon;
  • visualize a mathematical relationship;
  • create music or visual art;
  • analyze data;
  • model a system;
  • prototype a design;
  • test an idea;
  • communicate an explanation.

Technology, in other words, does not have to remain confined to the T in STEAM. It can become one of the instruments through which learners move among the disciplines, and AI can now participate in many of those same activities. The educational question is whether we treat that increasing capability as a reason to hand more thinking to the machine, or as an opportunity to deepen what the learner is doing.

Students as Teachers

This is where the First Five become particularly relevant to artificial intelligence. The student is not simply consuming technology; the student is teaching it. Teaching changes the relationship.

To make a computer do something, I first have to decide what I mean: identify the information that matters, articulate an instruction, establish rules or constraints, and examine the result.

When the result is wrong, incomplete, or unexpected, another question appears: Why? Was my instruction unclear? Was an assumption wrong? Did the computer interpret something differently than I expected? Or did its response expose something I had not fully understood myself?

The result is a deceptively simple learning cycle:

Question → Teach → Observe → Evaluate → Revise → Try Again

The computer’s response creates feedback, the feedback gives the learner something to evaluate, and evaluation creates another opportunity to learn.

A Connection I Recognized Later

Years after the experiences that produced the First Five, I recognized an interesting parallel with the work of Seymour Papert.

In Mindstorms: Children, Computers, and Powerful Ideas, Papert explored an educational relationship in which children program computers and encounter mathematical and other ideas through that interaction.3

The distinction in provenance is important:

Papert’s work is not the source of the First Five. The First Five originated in my own experiences teaching my first computer.

I include Papert because I later recognized a useful intellectual connection.

Both perspectives encourage us to look beyond the computer itself and consider what happens to learners when they construct, program, experiment, receive feedback, and make their thinking explicit.

For me, that produces another question:

What happens inside the learner when the learner has to make an idea clear enough for a computer to act upon it?

That question becomes even more interesting when the computer can respond in language.

From Output to Understanding

I returned to the First Five in 2025 while exploring natural-language processing. 4 The original activities had largely taught the computer to produce output: display text, generate sound, create visual forms, calculate values, and behave according to rules.

But output raised a deeper question. I had taught my computer to say “Hello World.” Had I taught it to understand what I was saying?

That question led me back through the other First Five. A computer could play a melody, but what would it mean to interpret meaning associated with music? It could draw something, but what would it mean to describe what the image represented? It could calculate, but what would it mean to reason about mathematical relationships? It could play according to rules, but what would it mean to adapt to strategy, context, or intent?

The question had changed from:

Can I make the computer do something?

to:

What does it mean for the computer to understand what I am trying to teach it?

Generative AI makes that question more important because increasingly fluent output can appear remarkably similar to understanding. That makes the human ability to question, interpret, and evaluate the system more important, not less.

Learning AI by Becoming the Teacher

In 2026, I returned explicitly to the First Five as an approach to AI literacy in The First Five Things to Teach a Computer: Learning AI by Becoming the Teacher. 5 The framework remained the same:

  • communication;
  • music;
  • art;
  • mathematics;
  • gaming.

What changed was my understanding of what happens to the learner. Teaching communication forces us to think about language, context, and meaning. Teaching music or visual creation introduces patterns, creativity, and expression. Teaching mathematics requires logic, structure, and abstraction. Teaching games introduces rules, strategy, feedback, and adaptation.

Across all five, the same principle emerges:

The act of teaching forces clarity.

The learner has to decide what matters: What is the objective? What information does the computer need? What assumptions am I making? What rules should remain fixed? What can change? How will I recognize a reasonable result?

The machine may be the pupil in the exercise, but the human is the one whose thinking has to become clearer. That is why I increasingly think teaching AI can itself become a learning strategy.

A Practical Learning Cycle

The First Five become most useful when they describe an activity rather than a list of topics. A teacher can apply the framework to almost any project by asking students to move through a repeatable cycle:

  1. Name the goal. What should the computer help create, calculate, represent, or explain?
  2. Teach the criteria. What information, examples, constraints, and definitions does the system need?
  3. Predict the result. What do students expect to happen, and why?
  4. Inspect the output. What is accurate, useful, surprising, incomplete, or misleading?
  5. Revise and justify. What changed, what evidence supports the change, and what remains uncertain?

This cycle makes the learning visible. A finished product is not the only evidence of success; students can also show their initial model, the instructions or rules they used, the errors they found, and the reasoning behind their revisions. In an AI-supported project, that record is especially important because a polished answer can conceal weak understanding.

The teacher’s role is not merely to approve the final result. It is to design conditions in which students must explain their decisions and, when appropriate, work without AI long enough to establish their own understanding. Students should also learn when not to use an AI system, what information should not be shared, how to acknowledge assistance, and when a human expert or primary source is required.

Where STEAM Meets AI

This creates an important distinction for AI education. AI does not necessarily need to become the educational destination; it can become another tool for learning across STEAM.

Imagine students confronting a meaningful real-world problem. Science might help them investigate it; mathematics might help them measure or model it; engineering might provide a process for designing and testing a response; and the arts might help them imagine alternatives, represent complexity, or communicate meaning. Technology might help them collect evidence, simulate possibilities, analyze information, create representations, or share what they have learned. AI can participate throughout that process, but the educational objective remains learning.

This is why I am cautious about reducing AI education to learning interfaces, features, or prompting techniques. Those skills may have practical value, but they are not enough.

The deeper question is:

Can students use AI as one tool among many to investigate, create, model, communicate, design, evaluate, and solve meaningful problems across disciplines?

That shifts our attention from AI proficiency toward human capability.

It also requires boundaries. AI readiness includes knowing that a system may be unavailable, inappropriate, biased, confidently wrong, or unable to protect sensitive information. A ready learner can choose a lower-tech method, seek another source, or decline to automate a decision when the risks or learning goals call for it. Readiness is therefore not measured by how often a student uses AI, but by whether the student can make a defensible choice about its use.

From STEAM to Systems Thinking

The more I worked with the First Five, the less they looked like separate categories. Communication influences games; mathematics underlies music and drawing; games depend on rules, probability, feedback, and adaptation; and creativity operates within constraints. One domain continually spills into another. That is systems thinking.

And it broadens the educational objective again. Real problems do not conveniently arrive labeled SCIENCE, MATHEMATICS, ENGINEERING, ART, or TECHNOLOGY. They arrive as systems.

People are part of those systems. So are technologies, institutions, infrastructure, economics, culture, information, policies, incentives, relationships, and values.

Artificial intelligence increasingly becomes another component.

The challenge, then, is not merely teaching students how to use AI. It is helping them reason about systems that include AI.

That realization has led me to a principle I now use in the Compass work:

We are not teaching students to use computers. We are teaching them to think in systems that include AI.

That may be the most important bridge between STEAM and AI readiness.

Curiosity Before Answers

There is another reason I like positioning students as teachers: teaching invites curiosity. Once learners discover they can teach a computer something, another question follows naturally: What else can I teach it? Curiosity creates experimentation, experimentation produces feedback, and feedback provides an opportunity to revise assumptions and try again.

The learning loop expands:

Question → Teach → Observe → Evaluate → Experiment → Revise → Learn

That cycle matters in an era when AI can produce an answer almost instantly, because an answer is not necessarily understanding. The learner still has to ask:

  • Does this make sense?
  • What assumptions are embedded in the answer?
  • What evidence supports it?
  • What might be missing?
  • How could I test it?
  • What changes if one of the conditions changes?
  • Can I explain why I think the result is reasonable?
  • When should I not use AI for this problem?

These are not simply prompting skills; they are habits of inquiry and judgment. Technology can accelerate parts of learning, but it should not eliminate the learning we were trying to accelerate in the first place.

From Evaluation to Judgment

Teaching a machine eventually introduces another responsibility: someone has to decide what the system should do. What should it create? What information should it use? What constraints should apply? Who might be affected when the system gets something wrong? Those questions move us beyond technical proficiency.

I now see the learner’s relationship with intelligent technology as an evolving one:

  1. User — I can operate the technology.
  2. Teacher — I can explain what I want the system to do.
  3. Evaluator — I can judge what the system produces.
  4. Designer — I can change instructions, constraints, and workflows.
  5. Systems thinker — I can understand the technology in relation to people and other systems.
  6. Steward — I accept responsibility for deciding how, when, and why the technology should be used.

This progression is my current synthesis, not a standardized AI competency model. But the distinction matters. As AI becomes more capable, knowing how to operate the technology may become less distinctive. Knowing how to question its output, evaluate evidence, recognize limitations, understand consequences, and exercise judgment about its use becomes more important.

That brings me to AI readiness.

What Does It Mean to Be AI Ready?

I increasingly think AI readiness is too important to define simply as access to AI or proficiency using an AI application. Those may be enabling conditions, but they are not the larger educational objective.

UNESCO’s 2024 AI Competency Framework for Students provides a useful external comparison. It organizes student competencies around four dimensions: a human-centered mindset, ethics of AI, AI techniques and applications, and AI system design. Its progression moves through Understand, Apply, and Create, and UNESCO describes students as responsible users and co-creators of AI.6

That framework was developed independently of the First Five. I include it because its emphasis on human agency, judgment, creation, and responsibility provides a useful point of comparison for the direction in which my own thinking has developed.

For me, the question becomes:

Can learners think, learn, create, evaluate, adapt, and exercise judgment in environments where intelligent machines are increasingly part of the system?

That produces a progression I find useful:

Technology as a Tool

What can technology help me learn or accomplish?

↓

STEAM

How can different ways of knowing help me investigate, create, and solve problems?

↓

Students as Teachers

What do I learn when I have to explain something clearly enough for a machine to act upon it?

↓

Computational Thinking

Can I decompose, sequence, model, abstract, test, and iterate?

↓

Systems Thinking

Can I recognize relationships, constraints, feedback, interactions, and consequences?

↓

AI Literacy

Can I understand and critically evaluate my interactions with artificial intelligence?

↓

AI Readiness

Can I inquire, create, reason, evaluate, adapt, and exercise judgment in systems that include AI?

AI readiness, understood this way, is not simply the next technology skill. It is a capacity for continuing to learn as technology changes.

One practical test follows from this definition: after using an AI system, can a learner explain the problem, verify the important claims, identify uncertainty, describe the system’s contribution, and defend the final decision? If the answer is no, the tool may have completed a task without deepening the learner’s understanding.

Turning the Teaching Machine Around

That brings me back to the relationship at the center of the First Five. Educational technology has long asked how machines might help teach people. Increasingly capable AI systems make that question more interesting than ever.

But the First Five invite us to reverse it:

What can a student learn by teaching a machine?

That question preserves human agency. It makes room for STEAM because meaningful problems cross disciplinary boundaries; computational thinking because teaching machines requires specificity; systems thinking because technology operates within relationships, constraints, and feedback loops; curiosity because unexpected results invite investigation; and ethics because people ultimately have to decide what these systems should do, for whom, under what conditions, and toward what purpose.

The tools will continue to change. My first computer required BASIC commands. Today’s students can interact with systems that generate language, images, software, analysis, and ideas. Tomorrow’s systems will be different again.

That is exactly why technology cannot be the ultimate educational objective. We need to develop human capabilities that endure as the tools change:

  • curiosity
  • creativity
  • communication
  • reasoning
  • evaluation
  • systems thinking
  • judgment
  • responsibility

The intellectual thread connecting this journey now seems surprisingly simple:

Technology is the tool. STEAM provides an interdisciplinary environment for learning with that tool. AI makes the tool increasingly intelligent and interactive. AI readiness is the human capacity to use intelligent tools with understanding, creativity, systems thinking, judgment, and purpose.

AI readiness is not readiness to hand thinking over to a machine. It is readiness to keep learning, thinking, creating, and exercising judgment in a world where intelligent machines have become part of the system.

That is where an original framework that began with five things I taught my first computer has taken me. And perhaps that was the lesson all along.

Further Reading

The First Five is my original framework. These writings document its development from my early programming experiences toward computational thinking, natural-language processing, and AI literacy.

Kronemeyer, Jason. “The First 5 Things to Teach a Computer.” Jason Kronemeyer, 30 Aug. 2024, www.jasonkronemeyer.com/education/computer-science/programming/2024/08/30/first-5-things-to-teach-computer.html.
The original published articulation of the First Five: communication, music, art, mathematics, and gaming. It also introduces the student-as-teacher relationship and identifies the five as building blocks of computational thinking. 1

Kronemeyer, Jason. “Teaching Computers how to Communicate: spaCy, NLTK, and Stanford’s NLP Textbook.” Jason Kronemeyer, 4 Nov. 2025, www.jasonkronemeyer.com/ai/nlp/education/tools/2025/11/04/NLP-Spacy.html.
A later derivative of the First Five that develops the distinction between producing output and interpreting human communication. 4

Kronemeyer, Jason. “The First Five Things to Teach a Computer: Learning AI by Becoming the Teacher.” Jason Kronemeyer, 23 Mar. 2026, www.jasonkronemeyer.com/2026/03/23/ai-connect-curiosity.html.
Extends the First Five into AI literacy and explicitly develops learning by teaching, curiosity, systems thinking, iteration, and the changing relationship between learner and intelligent system. 5

The following works did not originate the First Five. They provide external intellectual context for ideas that intersect with the framework as it has developed.

Papert, Seymour. Mindstorms: Children, Computers, and Powerful Ideas. Basic Books, 1980.
A useful comparison for thinking about learners programming computers and encountering powerful ideas through that interaction. Internet Archive. 3

UNESCO. AI Competency Framework for Students. By Fengchun Miao, Kelly Shiohira, and Natalie Lao, UNESCO, 2024, unesdoc.unesco.org/ark:/48223/pf0000391105. https://doi.org/10.54675/JKJB9835.
A contemporary external framework emphasizing a human-centered mindset, ethics, AI techniques and applications, AI system design, and progression from understanding to application and creation. 6

Notes

  1. I first publicly articulated the First Five framework in The First 5 Things to Teach a Computer on 30 Aug. 2024. The article identifies the five as communication, music, art, mathematics, and gaming; describes them as things I taught my first computer; places students in the role of teachers of their digital pupils; and connects the five to computational thinking. ↩ ↩2 ↩3

  2. STEAM is used here as a later interpretive lens for understanding the cross-disciplinary qualities of the First Five. It is not presented as the source of the framework. ↩

  3. Papert’s Mindstorms was published in 1980 and explores children programming computers and encountering mathematical and other ideas through that relationship. My comparison with Papert is retrospective and convergent. The First Five was not derived from Papert’s work. ↩ ↩2

  4. My 2025 article Teaching Computers how to Communicate explicitly builds on the earlier First Five and examines the movement from making computers produce output toward questions of understanding human communication. ↩ ↩2

  5. My 2026 Learning AI by Becoming the Teacher article explicitly returns to the First Five and develops their application to AI literacy. It describes the learner as an active participant rather than merely a user of AI tools and develops the idea that teaching forces clarity. ↩ ↩2

  6. UNESCO’s AI Competency Framework for Students was published in 2024. UNESCO describes 12 competencies across four dimensions and three progression levels: Understand, Apply, and Create. I use it here as a contemporary external comparison, not as a source for the First Five. ↩ ↩2