CodeKit™
An AI programming tutor for UK schools delivering personalised computer-science instruction and reducing teacher support time.
A computer-science teacher cannot debug thirty students at once.
Computer-science teaching in UK schools runs into the same wall every September. A class of thirty students, ten environments breaking in different ways, one teacher walking the room trying to keep everyone unblocked. The lesson plan slips, the strongest students wait, the weakest students drift, and the teacher's energy goes into syntax errors instead of teaching the concepts.
Most of the support burden is not pedagogical. It is environment setup, version mismatches, syntax typos the student cannot spot, and the kind of one-on-one debugging that does not scale across a classroom or a school week.
- Students must be able to run code without installing anything.
- The AI tutor coaches and explains; it does not write the answer for the student.
- Teachers retain control of the curriculum and authoring.
- Progress is visible per student so the teacher can intervene where it matters.
Inside CodeKit.
A patient coach grounded on the curriculum
The tutor walks the student through the concept, gives hints when they get stuck, and never just writes the answer. Grounded on the school's curriculum so the language matches the lesson plan.
No setup. No installs. Just open and start coding
A Python environment that runs inside the browser. The student writes code on one side, sees output on the other, and the teacher does not spend half the lesson fixing local installs.
Teachers author the curriculum; the AI does the rest
Lessons and exercises authored by the teaching team, organised by topic and difficulty. The AI tutor uses the same library the class is working from, so the help is consistent with the lesson.
Exercises with guided hints and self-checking
Every exercise is self-checking and gives the student a hint, then a stronger hint, then a worked example only as a last resort. The student stays engaged with the problem instead of jumping to the answer.
Teacher dashboard with per-student progress and gaps
The mission log shows which students are flying through, which are stuck, and which topics the class as a whole is finding hard. Targeted intervention instead of guesswork.
An AI tutor that coaches the student while the teacher keeps the classroom.
CodeKit pairs an in-browser Python environment with an AI tutor grounded on the school's curriculum. The student writes and runs code without local setup, asks the tutor for help when they are stuck, and gets a hint scaled to where they are, not the whole answer.
Teachers stay in control of what the class is learning. They author lessons and exercises, set the pace, and use the mission log to spot which students need attention and which topics the class is finding hard. The AI does the one-on-one coaching that does not scale; the teacher does the teaching that does.
- Curriculum-aligned AI tutor with hint-laddering: small hint, bigger hint, worked example as last resort.
- In-browser Python sandbox so a student opens the browser and codes immediately.
- Lesson and exercise authoring tool for the teaching team.
- Mission log per student with topic-level mastery and intervention prompts.
Teacher support time down 70%, computer-science enrolment up 45%.
Teacher time recovered
Roughly 70% less teacher support time inside lessons. The class moves at the pace the lesson plan intended.
Enrolment up
45% more students opting into computer science. A subject with a setup-friction reputation became a subject students chose because the friction was gone.
Personalised pace
The strongest students stop waiting for the slowest. The slowest students get help that is patient and specific without monopolising the teacher.
Visible progress
The mission log surfaces topic-level gaps across the class so intervention is targeted instead of guessed.
How we delivered it.
Stack
Capabilities
Compliance
From scoping to live.
- DiscoveryWorked with computer-science teachers to map where students get stuck and where teacher time goes. Most support time was syntax errors, environment setup, and one-on-one debugging that did not scale. Month 1
- Pilot buildAI tutor grounded on the curriculum, with an in-browser Python sandbox so students write and run code without environment setup. Months 2-4
- Production rolloutLesson library, exercise library, and mission-log progress tracking. Teachers author content; the AI tutor coaches the student through it. Months 5-6
Read another story
All case studies
Bring your team's next AI project to a 30-minute call.
No deck. We listen, sketch a delivery shape, and tell you honestly whether AI is the right tool for the problem.
