Three weeks in, the phone is face down on the bathroom counter and the brush is running its two minutes on whatever mode it was last left in. Dyson has built for exactly that: the CameraJet pairs over Bluetooth and appears in MyDyson, and it does not need the app opened to work [9]. So the component doing the expensive part of the job is also the one its owner has the least reason to look at.
Do the arithmetic on the capture rate and the response window and the system has roughly 2.8 frames to decide a gap is a gap and fire [16]. Loops that tight live or die on training data. Replayed at the camera's own capture rate, the whole image corpus comes to about four hours and forty minutes of footage [17], which makes the load-bearing number the six years of collection rather than the count of pictures [4]. That is a narrow, slowly assembled set, which is the honest way to do a narrow vision task and also the reason a competitor cannot answer it in one season.
Separate the thing being pitched from the thing being done. Pitched: optical gap targeting [3]. Done: a sonic brush with variable oscillation angles [19] carrying 12.5 ml of alcohol-free rinse, topped up at the dock in three seconds [12]. The mode named in the small print reads as the brush working without the jet firing, so the camera did not earn the headline plaque figure [6]. And that figure was measured against a proxy plaque Dyson developed with the National University of Singapore's Faculty of Dentistry, adopted so designers could iterate without commissioning a clinical trial after every change [7]. Good engineering economics. Thin consumer proof.
The buyer here is quite specific: someone who already owns a premium electric brush and a countertop water flosser, and who uses the flosser about twice a week. Dyson has form in reaching that person, in stick vacuums and premium hair dryers [15], and less happy form in two-in-one hybrids, given what Wired calls the short-lived Dyson Zone headphones [20].
The forcing function for anyone shipping a model the user cannot see: write down the verification event and its lag, then write down the in-product proxy that has to hold the user's belief until verification arrives. Here the verification event is a hygienist, six months out. Two proxies are already built: a live camera view of a jet hitting a gap, and coverage mapping with per-session feedback [10]. The feed is the better demo and the weaker proxy, since I expect it gets watched twice and then abandoned, while a coverage map at least reports whether the person's own brushing changed. I would make the map the default screen after a clean and treat the live view as onboarding theatre, and the cost of that is real: the live view is the one thing that visibly proves the hardware premium, and the announcement material in front of us does not carry a price [18], so nobody outside Dyson yet knows how much weight that proof has to bear.
