An AI Python tutor for a school.
Wolsingham School · an AI programming tutor for a one-teacher computing department
A student stuck on a Python exercise gets coached through it. The tutor never hands over the answer. It works from the lessons the school’s own teachers write, so the help matches what the class was taught. The school used it in class with Year 10 and below.
Overview
CodeKit began as pro bono work for Wolsingham School in County Durham, where OpenKit has held enterprise adviser roles for years. The school has one IT teacher and limited access to computing equipment, which makes offering computer science at GCSE something it has to build toward. The gap is resourcing.
So we built an AI tutor for learning to code. A student gets Python exercises with a tutor sitting alongside them that reviews what they have written and tells them how to improve it. The design rule was that it coaches, because a tool that writes the solution teaches a student to ask it for solutions. We aimed it at Year 10 and below, and the school used it in class.
Pro bono
Built for Wolsingham School in County Durham
One teacher
The whole computing department, with limited access to equipment
In the browser
Nothing to install on a school machine
Year 10 and below
The classes the school used it with, in lessons
DfE bid
CodeKit became the basis of the application, and we won the first phase
Challenge
Where the teaching time actually went
What stops a computing lesson working is usually practical: a machine that will not run the interpreter, or a typo the student cannot see. A full department absorbs those problems between its staff, and with one teacher they become the whole lesson.
Approach
Inside CodeKit
OpenKit built CodeKit against that constraint. It runs in the browser, so nothing needs installing on a school machine, and it gives every student something patient to ask when they get stuck. The curriculum stays the school’s, and the teacher gets a view of where the class is struggling so their attention lands where it counts.
Python that runs in a school browser
The student opens a tab and writes code, with the output next to the editor. There is nothing to install and no version to match, so thirty machines do not have to reach the same state before a lesson can begin on the first exercise.
Exercises the school authors
The school’s teachers write the lessons and exercises and organise them by topic and difficulty, and the tutor works from that same library. Help that contradicts the lesson is worse than no help, so the tutor never gets to invent its own curriculum.
Where the class is stuck
The mission log shows per-student progress and topic-level gaps across the group, so a teacher covering everything on their own can see where to spend the attention they have.
The build
What CodeKit is built from
- AI tutor agent grounded on the curriculum
- In-browser Python sandbox
- Lesson and exercise authoring tool
- Mission-log progress tracking
Controls on this project
- Operates to UK GDPR
- UK data residency
OpenKit certifications
- ISO 27001
- ISO 9001, UKAS-accredited
- Cyber Essentials
The school
Working with the school beyond the software
- Lessons delivered in person at the school, taught by the people who built the tutor.
- Career days, so students hear what the work looks like from people doing it.
- Enterprise Adviser roles held at the school through the regional partnership, over several years.
Result
What CodeKit led to
When the Department for Education opened a competition for AI tools in education, the bid started from work already done. We had a subject-specialised assistant that reviewed student work and explained itself in language a fourteen-year-old could use, a school that knew us, and a working answer to the question every education buyer asks first, which is whether the thing does the child’s homework for them. That experience and that codebase became the basis of the application, and we won the first phase.
Other engagements
What could your business do with AI?
Find out with experts who become part of your team. We uncover opportunities, get ideas working, and help your people build on the results. We reply within one working day.
Get in touch AI Engineering