The school that came out of a robot.
Catalyst exists because a group of school students reached the semi-finals of the ANA Avatar XPRIZE, and we realised — watching them work — that we had been doing school wrong for years.
The robot that changed our minds.
The ANA Avatar XPRIZE was a global competition with a $10M prize pool, asking teams to build fully teleoperated avatar systems — robots a human could operate remotely, with vision, hearing, manipulation, and a sense of touch. The serious field was university research groups and industrial labs.
We were a school. We entered anyway. Our students designed and built the avatar themselves over two years — the aluminium frame, the vision pipeline, the haptics, the batteries. When the qualifying round finished, two teams from England had made it through to the semi-finals: a team from the University of Sheffield, and us. COVID stopped us from going further.
What stayed with us wasn't the placing. It was watching what those students could do — and realising the curriculum we were supposed to be delivering had almost nothing to do with any of it. The two-hour written paper, sat alone in a hall, on a topic the student would never revisit: that wasn't preparing anyone for the kind of work we were doing in the workshop. It was the wrong measurement of the wrong thing.
Catalyst is the school we built so the next group of students could start where those students finished.
What the students built
- PlatformFully teleoperated avatar
- VisionOpenCV on Raspberry Pi with Intel Neural Compute Stick
- HapticsCustom feedback system, designed in-house
- FrameAluminium, machined on-site
- PowerIn-house Li-ion battery packs
- Stage reachedXPRIZE semi-finals (2021)
Why we walked away from exams.
The traditional school qualification was designed for a world that hasn't existed for a generation. It rewards a specific skill — reproducing facts under timed conditions, with no tools — and treats that one skill as a proxy for capability. It isn't. Working life looks nothing like that exam hall. The students who do well in their twenties are not the ones who memorised faster; they are the ones who can scope a problem, gather the materials, build something that works, and explain every choice in it under questioning.
So we don't run GCSEs. We don't run A-levels. We don't run mock exams or revise toward grade boundaries. We build, we ship, we defend. Every term, every student, every year. That's the school.
The four things we hold to.
Make, don't just sit
Every term, every student ships something — a model, a paper, a build, an analysis, a film. Real artefacts, defended in person.
Subject mastery, first
AI is a force multiplier, not a substitute. Students cannot direct what they don't understand, so deep knowledge of maths, science, language and history remains non-negotiable.
Human teachers, amplified
AI handles marking, retrieval practice and intervention triage. Teachers spend their time where it matters: coaching, challenging, knowing the child.
International by design
A global student body, taught with global examples, prepared for a labour market that doesn't stop at any border.
How we think about AI in lessons.
We publish our principles openly because families deserve to know what their children are being taught and how. These are the rules we hold ourselves to.
Principle 1
AI is a tool, not an author. Students must be able to explain, defend and reproduce any work they submit. Viva-style checks are routine, not exceptional.
Principle 2
Provenance is recorded. When AI contributes to a piece of work, students log the prompt, the model and what they changed. Honest use is graded; hidden use is treated as plagiarism.
Principle 3
Human in the loop, always. No academic decision — grading, progression, intervention, reporting — is taken by an AI alone. A teacher signs off every time.
Principle 4
Safeguarding first. Tools are vetted, logged and age-gated. We use enterprise-grade systems with data residency in the UK and EU, and we do not train external models on pupil data.
Principle 5
Bias, error and uncertainty are part of the syllabus. Students learn how models fail, where they hallucinate, and how to audit them — because the world will need people who can.
Principle 6
Screen time is purposeful. Discussion, writing by hand, lab work, workshop time and reading remain core. AI extends those activities; it does not replace them.
Faculty.
Our teachers are recruited for two things: deep subject expertise and a working command of AI as a teaching tool. Many of them were on the XPRIZE project. Every member of staff completes our internal certification in AI-enabled pedagogy in their first term, and continues professional development each year. Subject leads include engineers and researchers who have shipped real systems, not just taught theory.
See how to apply