Exact Science
Get started
ProgrammingWeekly course

AI Foundations

Learn how modern AI actually works: the maths and code underneath, then build projects that show judgement, not just prompts.

About this course

A project-based course for ambitious GCSE and A-Level students who want to understand modern AI properly: the maths and coding underneath, not only how to use a chat box.

Over eight weeks you will ship three portfolio pieces: a Verification Notebook, a Subject Study OS, and a Make & Defend artefact. You will build intuition for representations, probability, training vs inference, and simple code labs, then apply that judgement to real tools.

Format

8 weekly sessions × 90 minutes, small group (up to 6). Homework is 45–60 minutes of maths, code, or project work tied to that week's idea.

What you leave with

three shipped projects, a working mental model of how these systems behave, and the habit of checking claims yourself.

This course is not

exam cheating techniques, “write my coursework”, tip-sheet prompting, or treating AI as a substitute for subject mastery.

Weekly plan

Project-based

Project map

  1. 1

    Verification Notebook

    (weeks 1–2): claims checked against maths, code, or primary sources, then re-explained without the model.
  2. 2

    Subject Study OS

    (weeks 3–5): a revision system for one GCSE or A-Level subject, built with clear honesty rules and an explain-back loop.
  3. 3

    Make & Defend

    (weeks 6–8): one shipped artefact (small model experiment, data analysis, researched briefing, or creative piece) plus a viva.
1

Week 1

What these systems are (and are not)

Tools
one frontier chat model; pen and paper.
In class
tokens, context, hallucination; training vs inference in plain language; live failure clinic on familiar syllabus topics.
You ship
start your Verification Notebook with five checked claims from maths or science.
Project checkpoint
P1 start.

Teacher notes

Session goal: replace magic with a mental model. Demo failure: confident wrong derivation. Exit: notebook template live with five checks underway.
2

Week 2

Representations, probability, and precise briefs

Tools
chat + simple Python or spreadsheet for tiny probability checks.
In class
how inputs become representations; sampling and uncertainty intuition; prompting as problem specification (goals, constraints, examples).
You ship
Notebook: before/after briefs plus an independent maths or code check.
Project checkpoint
P1 ship.

Teacher notes

Session goal: link brief quality to verification. Demo failure: vague prompt that invents a “proof”. Exit: P1 complete with at least one fully verified before/after pair.
3

Week 3

Learning with models, not outsourcing thinking

Tools
chat + notes app; optional short code notebook.
In class
Socratic tutoring patterns; spaced retrieval; mark-scheme aware revision without copying; when a model is a coach vs an answer key.
You ship
Study OS outline for one subject plus honesty rules.
Project checkpoint
P2 start.

Teacher notes

Session goal: AI as coach. Demo failure: model that “does the homework”. Exit: subject chosen and honesty log written.
4

Week 4

Data, sources, and research hygiene

Tools
research-mode tools; primary sources; simple plots in Python or sheets.
In class
citation hygiene; spotting confident nonsense; reading a short data table or plot before trusting a summary.
You ship
source-backed mini-brief on a school topic with one checked figure or calculation.
Project checkpoint
P2 build.

Teacher notes

Session goal: every claim needs a trail. Demo failure: invented citation or misread chart. Exit: mini-brief with sources the student opened themselves.
5

Week 5

Writing, critique, and Study OS v1

Tools
chat + docs.
In class
draft → outline → red-team; keep your voice; turn Study OS into a quiz bank with explain-back.
You ship
Study OS v1 for one upcoming topic or paper.
Project checkpoint
P2 ship.

Teacher notes

Session goal: critique over polish. Demo failure: fluent prose that loses the student's voice. Exit: Study OS usable without the chat open.
6

Week 6

Making: code, data, or creative with foundations

Tools
Cursor/Copilot-style assistants or a small notebook; optional image tools.
In class
build a thin working slice; verify outputs with a test, plot, or hand check; choose capstone type.
You ship
capstone proposal + first working slice.
Project checkpoint
P3 start.

Teacher notes

Session goal: ship a thin slice. Demo failure: generated code that fails a trivial check. Exit: proposal accepted and slice demonstrated.
7

Week 7

Memory, projects, and limits of agents

Tools
Custom GPTs / Claude Projects / notebooks as personal systems.
In class
context and memory hygiene; when automation lies; mid-build critique.
You ship
capstone mid-build + failure-mode notes tied to the underlying model limits.
Project checkpoint
P3 build.

Teacher notes

Session goal: systems with guardrails. Demo failure: agent that invents progress. Exit: three failure modes written down.
8

Week 8

Capstone + integrity clinic

Tools
portfolio of all three projects.
In class
present; viva on maths/code decisions; exams vs coursework vs life.
You ship
final Make & Defend artefact.
Project checkpoint
P3 ship.

Teacher notes

Session goal: defend the work cold. Demo failure: cannot explain a decision without the chat log. Exit: three projects shipped; clarity of thinking over flashiest tool use.

We'll call or email you back

Mon–Sun, 8am–9pm

Questions?

Get in touch

Ask about this course or anything else. We will help you choose what works best for you.