Introduction
Most teachers using AI are doing the same thing: opening ChatGPT, typing "write me a worksheet on fractions," and copying the output into a Word document. That is useful. It saves thirty minutes. But it barely scratches the surface of what is now possible.
In early 2025, AI researcher Andrej Karpathy coined the term "vibe coding" — a new approach to software development where you describe what you want in natural language, and AI agents write the code for you. You do not need to be a programmer. You need to be clear about what you want.
I have been using this approach — with AI coding tools like Antigravity, Cursor, and Claude Code — to build professional-quality teaching resources at scale. Not generic worksheets. Not rough drafts that need heavy editing. Complete, formatted, curriculum-aligned exercise booklets generated programmatically from structured data.
The results have been transformative. And the barrier to entry is far lower than most teachers assume.
What Is Vibe Coding for Teachers?
The difference between traditional AI use and vibe coding is this:
Traditional approach: "Write me a worksheet on persuasive writing for Year 9." The AI generates a generic worksheet. You format it manually. You tweak the questions. You print it. It is fine. It took you fifteen minutes instead of forty-five.
Vibe coding approach: you provide the AI with structured context — your semester plan, your curriculum objectives, your student performance data — and ask it to write code that generates resources programmatically. The output is not a single worksheet. It is a system that produces worksheets, exercise booklets, vocabulary lists, comprehension tasks, and answer keys — all formatted, all aligned to your curriculum, all ready to print.
The key difference is that you are not asking for content. You are asking for systems that produce content.
This is a fundamental shift. Instead of generating one worksheet at a time, you build a pipeline that can generate an entire term's resources from a single set of inputs.
How It Works in Practice
Here is the workflow I use:
Step 1: Create a markdown file with your semester plan. This includes units, learning objectives, key vocabulary, assessment criteria, and pacing. Markdown is just a simple text format — no special software needed. It looks like a structured list.
Step 2: Add CSV files with student performance data. Scores from recent assessments, identified gaps, areas for improvement. A simple spreadsheet exported as CSV.
Step 3: Feed these files as context to an AI coding agent. Tools like Antigravity (built by Google DeepMind), Cursor, or Anthropic's Claude Code can read these files and use them as context for code generation.
Step 4: The agent writes scripts that generate resources. The AI creates Python or Node.js scripts that programmatically generate Word documents, PDF exercise booklets, images, or slide decks — all aligned to the curriculum objectives and tailored to student needs.
This is not theoretical. I have used this workflow to create 30-page weekly exercise booklets in English, aligned to specific IGCSE units, with auto-generated answer keys, differentiated tasks, and consistent formatting. What would have taken an entire weekend of manual work was completed in an afternoon.
Why Code-Generated Resources Are Better
Consistency
Every worksheet, every exercise booklet, every assessment follows the same design template. Fonts, spacing, branding, header formatting — all standardised across the entire set of resources. The result looks professional, not cobbled together.
Scalability
Regenerating resources for a new cohort, a new unit, or a different curriculum takes minutes, not hours. You change the input files — the semester plan, the vocabulary list, the student data — and rerun the scripts. A new, customised set of materials appears.
Data-Driven Differentiation
This is where the approach becomes genuinely powerful. Student performance data feeds directly into resource generation. A student struggling with subject-verb agreement gets targeted grammar practice. A student excelling at comprehension gets extension reading tasks. The differentiation is not teacher-estimated — it is data-driven and specific.
Version Control
Resources are generated from source files — the markdown plan, the CSV data. This means every version is tracked and reproducible. When you update the plan for next year, you can see exactly what changed. When you share resources with a colleague, they can regenerate them from the same source files with their own adaptations.
This is how software teams work. And there is no reason teaching teams should not work the same way.
Getting Started: You Don't Need to Know How to Code
The entire point of vibe coding is that you do not need to code. You describe what you want — "generate a 20-page exercise booklet on persuasive writing, with a vocabulary section, three comprehension tasks per chapter, and an answer key at the back" — and the AI writes the code to make it happen.
You review the output. You refine your instructions. You iterate. The AI handles the technical implementation.
Tools like Edcafe AI and Code.org are making AI-powered resource creation more accessible to educators. But the vibe coding approach goes further — because you are not limited to pre-built templates. You can build anything you can describe.
The learning curve is real but manageable. Start small — a single worksheet generator. Then build up to full booklets. Then add student data pipelines. Each step builds on the last, and the productivity gains compound rapidly.
Conclusion
AI in education is stuck in "ChatGPT prompt" mode. Teachers type a request, get a generic response, and spend almost as long editing the output as they would have spent creating the resource from scratch.
Vibe coding unlocks the next level: automated, scalable, curriculum-aligned resource generation. It treats teaching resources not as one-off documents but as outputs of a systematic, repeatable process.
Teachers who learn to work with AI coding agents will save hundreds of hours per year. They will produce higher-quality, more consistent, more differentiated materials than they ever could manually. And they will have more time for the work that actually matters — planning brilliant lessons and building relationships with students.
Have you tried using AI to generate teaching resources? Are you using code-based tools, or sticking to ChatGPT prompts? What has worked and what hasn't? Share your experience in the comments below.
Comments (0)
No comments yet. Be the first to share your thoughts!