Four University of Canterbury engineering students have built a machine that solved a 4×4 Rubik’s Cube in 9.806 seconds — a time that, if confirmed by Guinness World Records, would beat the existing robotic world record of 45 seconds by a factor of nearly five, and the fastest human record of about 15 seconds as well.
The robot’s official record attempt was made on Friday 16 October at UC, with confirmation from Guinness still pending as of Monday. The team — final-year mechatronics students Caleb Everitt, Hayden Watson, Maz Cameron and Johann Torres — turned a final-year project into a year-long chase of a record that had looked untouchable in both directions.
Why the old robot was slow enough to beat
The spark was a social video of the previous record-holding robot taking 45 seconds to solve the same cube — roughly three times slower than the human record. That gap was the tell. Humans solve with two hands and full cube mobility; most solving robots clamp the cube and rotate it face-by-face, paying a reorientation tax on every move.
The UC machine attacks that tax directly. According to UC’s own release, four cameras read the cube’s colours, an open-source solving algorithm computes the move sequence, and the mechanical design lets the mechanism access all three cube axes at once — eliminating the reorienting steps that slowed the previous holder. Faster motors do the rest.
The unglamorous 90 percent
What is instructive in the team’s account is how little of the year was spent on the puzzle itself. Per The Gist’s summary, the hard problems were classic physical-AI labour: distinguishing colours under varying workshop lighting, and stopping the mechanism jamming at solve speeds. The fixes — refined software, stiffer structural rigidity, suction cups for precision movement — are the same iterate-and-harden loop that separates a demo robot from a reliable one.
Division of labour also mattered: each student owned one layer — Everitt mechanical, Torres electrical and motor control, Watson computer vision, Cameron software. That four-layer split is a miniature of how any production robotics team works.
The NZ angle
For a country whose robotics coverage skews imported — Boston Dynamics warehouse footage, US humanoid funding rounds — a Christchurch-built machine setting a pending world benchmark is the relevant local signal: the full stack from cameras to solver to mechanism was designed and assembled in NZ, by undergraduates, on a university bench. As we noted when NZ was told it is already behind on deep AI integration, the constraint story keeps being that capability exists here but stays under-funded and under-visible. A record attempt is the cheapest possible counterexample.
UC’s engineering pipeline feeds exactly the firms building the country’s agricultural and industrial autonomy — a sector we have covered scaling fast, albeit mostly offshore. Whether the university converts a viral record into deliberate robotics funding is the follow-up that matters; the precedent elsewhere is that attention without funding dissipates in a news cycle.
The team is awaiting Guinness verification. Until that lands, the honest framing is: a verified 9.806-second solve on camera, a documented engineering process, and two records in reach.
Sources: Inside Government NZ · The Gist NZ · University of Canterbury
