Christ Church Grammar School

Alex Coulter

Robotics Learning Practitioner

For years the biggest barrier to student projects in my classroom hasn't been mechanical design, CAD modelling or even circuit wiring. It's been the coding. A student could model a robot chassis for manufacturing, understand torque and gear ratios and correctly wire a motor controller and still hit a wall the moment they needed to write the program that brought it all together.

Coding from scratch demands a level of syntax fluency and logical structuring that takes most students months, sometimes years, to build and that's time a single term or semester-long project doesn't allow. So, much like the professional developers before them, my students turned to what was available: forums, repositories, old project write-ups. But these examples were often narrow and specific to someone else's build, using a particular sensor or a particular motor driver and translating that logic to their own project required exactly the coding confidence they didn't yet have. Too often a promising mechatronics build stalled not because the idea was flawed, but because the student couldn't get past the coding hurdle.

With the introduction of AI coding hasn't become unnecessary, it's become accessible. Despite its often-negative press and reputation in the media, AI has shifted the entry point for students: these tools can now generate the bulk of a working codebase, complete with pseudocode-style comments explaining what each block does and why. It's now possible for a student with little to no coding experience to look at a line reading “// read distance sensor and stop motor if object is closer than 10cm” and immediately understand not just what the code does, but why it's structured that way.

That's a fundamentally different starting point to teaching coding purely from first principles, line by line, which takes longer and produces noticeably less engagement. Students disconnect from the abstract exercise of "learn the language first, build later," particularly when what they're learning isn't directly relevant to what they want to build. When students are instead handed a working example tied to their own mechatronics project and asked to read it, annotate it and explain it back, something different happens. They're not memorising syntax in isolation, they're reverse-engineering a real solution to a problem they already care about. The code stops being a black box and becomes a design artefact, no different to a CAD model or circuit schematic: something to be pulled apart, understood and reshaped.

“Reverse-engineering code and re-designing a physical solution are, in practice, the same skill applied to two different mediums.”

This is where the "reprogramming" part comes in. Once students can explain the AI-generated code in front of them, they start adjusting it: changing threshold values, adding conditions, combining functions the AI wrote separately, stripping out what they don't need. That act of modification is where the real learning happens and it's the same instinct we already nurture in the workshop, where a student takes an existing bracket design and reworks it to fit their own mechanism. Reverse-engineering code and re-designing a physical solution are, in practice, the same skill applied to two different mediums.

The result has been a considerable shift in confidence that goes well beyond coding ability, which has led to improved student outcomes. Once students believe they can make the code do what they need, even if AI wrote the first draft, they stop seeing programming as the ceiling on what they can build. Ideas once quietly shelved because "the code would be too hard" are now being attempted. We've seen a marked improvement in the quality and complexity of student projects as a result: more sensors integrated, more complex responses programmed, more ambitious builds attempted overall. The work is more creative and technically sophisticated and reflects a deeper understanding of engineering and coding concepts, achieved without students needing extensive computer science experience or countless hours learning fundamentals first.

None of this replaces the value of understanding programming fundamentals and there's a real conversation to be had about where AI-generated code should stop and a student's own understanding should begin. But used deliberately, as a worked example to read, question and adapt rather than a shortcut to copy and submit, AI is removing what was, for many students, the single largest obstacle between having an idea and actually building it.