Mindset Our Focus Security Stories Get in touch
All stories
Stories/Programista AI/Curriculum
25 May 2026 · Curriculum

We wrote a whole new school subject, and half the grade is one real person's problem

"Wstęp do AI" is a new first-year subject for technik programista: seven blocks, 32 hours, and a full teaching pack behind it. The centrepiece is not an exam, it is a working prototype each student builds for someone they actually know.

Program nauczania · przedmiot autorski, v1 Wstęp do AI Technik programista (351406) · klasa I · semestr 1 · 32 godziny lekcyjne
BlokTematGodz.
1Czym jest AI, modalności, demo na żywo4h
2Jak AI naprawdę działa: tokeny, parametry, atencja6h
3Promptowanie: od naiwnego do mistrzowskiego6h
4Modele lokalne + MCP + asysta programistyczna6h
5Budowanie z AI: Spec Driven Development + mini-projekt5h
6Etyka i ekonomia AI: debata stakeholderów3h
7Sprawdzian kompetencyjny2h
Razem32h
The seven blocks, as they appear in the programme itself.

Most "AI in schools" initiatives end at a guest lecture. This one had to end at a subject a teacher can actually walk into and teach, with stated learning outcomes, how each one is verified, hours per block, what the lab needs, and how the year is graded. So that is what we wrote.

Half theory, half hands-on, deliberately

Thirty teaching hours split fifteen and fifteen. Every theory hour has a practical counterpart in the same or the next block: tokenisation on paper, then counting tokens for Polish versus English text; the Transformer diagram, then setting temperature to 0, 0.5, 1.0 and 1.5 and documenting what changed in the answers.

Students work in VS Code with Copilot from block one. Block four puts a small model on the classroom machine with Ollama and wires it to a local data source over MCP, so "cloud versus local" stops being a slide and becomes a comparison they ran themselves.

The mini-project is for one real person

Block five is worth half the year's grade, and it is not a topic from a list. The student finds someone in their life: a grandmother, a sibling, a friend, interviews them, and writes a specification from that conversation before any code exists. Then they build the prototype with an AI assistant, test it with that person at least twice, fix what the testing found, and record a five-minute video documenting the interview, the demo and one quote from the user.

We deliberately did not hand out a list of ten project ideas. A list would have killed the only part that matters: a fifteen-year-old noticing a real need in a real person. What they get instead is one worked example, a BMI calculator built for someone's grandmother, and the criteria for choosing their own.

Half the hundred points are for the code against the specification. The other half are for the process with the user: the interview, the two test sessions, the iterations visible in the commit history, the quote.

How the year is graded
Mini-project50%
Test after block 220%
Prompt portfolio10%
Stakeholder debate10%
Participation + self-assessment10%

Everything a teacher needs to run it

A programme on its own gets read once and forgotten. The package around it is thirty-odd files: lesson plans for all seven blocks, six student exercises with their answer keys, four grading rubrics, a short test after block two, the end-of-module competency exam, a terminology convention for Polish and English AI vocabulary, and a decision log explaining why the module is shaped the way it is.

One thing the module cannot fix: AI does not yet appear in the national vocational exam papers. The competency exam at the end of block seven is the school's own, it earns a module certificate and deliberately does not affect the annual grade.

More updates
← Earlier February 2026 · Workshop

AI handles the easy stuff. So every junior now starts where seniors used to be

Later → 8 June 2026 · Talk

The machine writes the code. The human decides

Back to all updates

Your story could be next

Book a discovery call