AI literacy in coding education means students understand how AI assistants like Replit Agent work, where their limits lie, and how to use them deliberately as a tool — rather than relying on them blindly. The goal is not to write less code, but to ask better questions and evaluate the output critically.
Why AI Literacy Now Belongs in the Curriculum
Swiss education frameworks — from compulsory school through vocational training — are increasingly embedding digital competence as a core subject. Lehrplan 21 establishes computer science and media as an independent area. What is still missing: concrete tools and methods for practising the use of generative AI.
Replit Agent is ideal for this because it does exactly what learners expect from AI tools — and makes the consequences of generated code immediately visible in the browser. Well-crafted prompts produce working apps. Poor prompts produce bugs that can be analysed together as a class.
According to the Swiss Federal Statistical Office (BFS), 61% of 15- to 24-year-olds in Switzerland already use AI assistants regularly for learning tasks in 2025. Schools that ignore this topic lose pedagogical relevance.
The Pedagogical Model: Prompt → Code → Analyse → Iterate
Phase 1: Write the Prompt (5–10 minutes)
Students describe in plain language what their app should do. No technical terms allowed — everyday language only. Example: "Create a website where I can enter my daily tasks and check them off."
Teacher's role: discuss prompt quality before Replit Agent is started. What is precise enough? What is missing?
Phase 2: Analyse the Code (10 minutes)
Replit Agent generates HTML, CSS, and JavaScript. The class examines the code together — with one central question: "What do you recognize that you already know?"
This connects AI output to previously learned material. Students see <button> and recall the HTML introduction. They see function and link it to the concept discussed last week.
Phase 3: Errors as a Learning Moment (10 minutes)
Deliberately give a flawed task: "Create an app that always knows what the user wants to do next." Replit Agent will produce either meaningless code or a very generic result. Discussion: Why? What can AI not know?
This builds a realistic understanding of AI: it is not a mind-reader or a universal solution — it is a tool with strengths and weaknesses.
Phase 4: Iterative Improvement (15 minutes)
Students take the generated app and improve it through targeted follow-up prompting: "Add the ability to delete completed tasks." Or: "Change the colors to match the school colors."
Every iteration is a micro-feedback loop: prompt → result → evaluation → next prompt. This exact cycle is a core skill in the 2026 workplace.
Five Concrete Lesson Ideas Using Replit Agent
1. Personal Timetable Generator (Middle school)
Task: "Create a website that shows my weekly class schedule as a table."
Learning objective: table structure, CSS styling, inserting own data.
2. Quiz App for Exam Preparation (Secondary school)
Task: "Build a quiz with five questions about [subject matter]. After each answer, the app should show whether I was correct."
Learning objective: logical branches, browser-side data persistence, UX thinking.
3. School News Feed (Vocational college / Gymnasium)
Task: "Create a simple website that displays school announcements. New announcements should appear at the top."
Learning objective: arrays, DOM manipulation, data structures.
4. CO₂ Calculator for the School Commute (Cross-curricular)
Task: "Build a calculator that computes CO₂ emissions for the school commute based on transport mode and distance."
Learning objective: forms, calculation logic, the social relevance of code.
5. Student Portfolio (Year-long project)
Task built iteratively: start with "Create my personal website", then extend week by week.
Learning objective: version control, project structure, professional self-presentation.
Replit Agent vs. Other AI Coding Tools for Education
| Tool | Strengths in the classroom | Weaknesses |
|---|---|---|
| Replit Agent | Complete app in seconds, live preview, compatible with Swiss data protection (EU servers selectable) | Requires basic prompt skills for meaningful results |
| GitHub Copilot | Deep IDE integration | Requires local installation |
| ChatGPT Code Interpreter | Good explanations | No live deployment, no visual code editor |
| Scratch AI extensions | Beginner-friendly | No real code, limited transferability |
Concerns About AI in the Classroom — and How to Address Them
"Students won't learn anything if AI writes the code."
This misframes the pedagogy. When the teacher focuses on analysis, iteration, and understanding, AI-generated code is learning material — not the answer. Compare: a maths teacher shows a solved problem and explains the solution path. The student then solves one themselves.
"We can't verify whether students did their own thinking."
Adjust the assessment format: instead of "deliver a working app" → "explain in three sentences what this code does and why you wrote it this way." A brief oral check directly after the coding task works well.
"Replit data is in the cloud — is that GDPR-compliant?"
Replit offers European server options. For Swiss schools, we recommend using exclusively Replit Teams for Education (no personal data, no content that implicates school law). CNEXT advises on the specific configuration.
Teacher Training: How CNEXT Supports Schools
CNEXT offers half-day workshops for teachers — at the school or remotely. Content includes:
- Hands-on first experience with Replit Agent
- Pedagogical framing: when to use AI, when not to
- Lesson planning: three ready-to-use sequences for immediate use
- Parent communication: talking points on the topic of AI in the classroom
Book an appointment — the first three workshops in the 2025/26 school year are free for pilot schools.
"AI literacy is no longer an additional skill — it is the foundational competence of the 21st century. Teachers who embed Replit Agent pedagogically give their students a decisive advantage — not because they use AI, but because they understand how."
— Marcel Haas, CEO CNEXT, Replit Master Builder Level 5

