Five days, fifteen manufacturers, the same routine every single day: leave the hotel at 9, grab a luckin coffee on the way, arrive at a vendor, sit down for tea and fruit, tour the facilities, look at samples, get more coffee, visit another vendor, then another, back to the hotel late. Repeat.
Most factories looked like normal office buildings from the outside. The good ones, you knew immediately - they have every machine running, every person moving, automations and dashboards on full display. The lower-tier ones you could spot just as fast: dust sitting on the equipment. In one factory, a room I walked through was dedicated to adhesives and nobody was wearing masks, despite the air being so thick it made you dizzy just passing through. Another, about ninety minutes outside Shenzhen where the rent gets cheaper and the regulations get looser, had its windows smashed in and a floor you could barely see under layers of newspaper clippings and grease. Gears were moving at full speed, inches from bare hands.
I didn’t ask about any of this. It was just “the floor”.
After every tour came the showroom. Walls lined with samples from small components, housings, PCBs, to full assemblies. Many of these factories had supplied Huawei or BYD. One made hardware for luxury brands like Celine. Occasionally someone would point at an enclosure and say Pingguo - Apple - like they were mentioning a mutual friend.
When I asked what other companies they were building a given part for, they told me, without hesitation. NDAs exist in Shenzhen technically, but in practice information moves very differently. One production line I walked past had been physically locked down, so that even employees from the same factory weren’t allowed in. It was the exception. Everywhere else, I was shown everything. Parts from competing products sitting in buckets at the end of the same line, some of them still hot from the injection molding process. The same machines, the same tolerances, the same hands, were building five competing products simultaneously before lunch.
When IP and trade secrets flow freely across the factory floor, the results are exactly what you’d expect.
Several manufacturers I visited were being actively sued by Oura Ring and Whoop. Not for cloning the exact product, but for the same sensors, the same health metrics, the same fundamental premise. Some of the competing devices were thinner than the Oura Ring and more technically sophisticated in terms of form factor. Oura’s legal theory relies on trade secret misappropriation - that proprietary algorithms and manufacturing tolerances were transferred through shared supply chains. Whether they’ll win is genuinely unclear; courts have struggled for years to determine where legitimate knowledge transfer ends and theft begins in a system designed to blur that line. What’s less unclear is the dynamic that produced the lawsuit: a supply chain optimized for speed and flexibility is, by design, not optimized for exclusivity. The component you thought was yours gets commoditized the moment a competitor shows up with the same RFQ.
Which brings us back to the AI wrapper problem, except here, the wrapper has been standard practice for decades. You can build a wearable today without designing one. Find the right factory, select a form factor already coming off their line, write software on top, make a logo, put it in a box. Six months, maybe less. Plenty of companies, including startups that have raised tens of millions of dollars, have done exactly this. The product exists, it ships, and it reviews fine. And it is, structurally, indistinguishable from what the factory will sell to the next person who walks in.
The honest version of the question this raises is not “is this defensible?” - most of it isn’t, legally or competitively - but “then what is?” Brand, yes but usually not on its own. Software and data, when the software is genuinely differentiated and the data is genuinely yours. Trust, if you’ve earned it and haven’t done anything to lose it. Garmin has survived a half-dozen disruption cycles not because its sensors are proprietary but because its maps are, and because it spent twenty years selling to pilots and soldiers and endurance athletes who don’t switch easily. Apple Watch won on ecosystem and five years of health data that now costs you something real to abandon.
The factories in Shenzhen are not going to solve any of that for you. They will build the object to whatever tolerance you specify, meet your deadline, match your cost target, and have a nearly identical object in production for a competitor within the quarter. What they cannot manufacture is the reason someone chooses your box over theirs. Or why, having chosen it once, they come back.
Wrappers aren’t just a software problem. Hardware has had the wrapper problem for years. The difference is you can't vibe-code your way out of a bad supply chain over a weekend. The companies that survived didn’t win on what they sourced, they won on what couldn’t be copied off a line.
We want Taya to be a product people love. If it is, it will get copied, that’s just how this works, and we’ve now seen it up close. Some of our components are sourced in Shenzhen. We’re assembling in the US for our first production run, because we want tighter control over quality, even though it costs more. The bet is that what the factories can’t replicate, from the care that went into the product to the trust we’ve built, is worth more than the savings. We’ll find out.
- Elena
Join the Journey
CRAZY!
The US assembly call is the right one, and not just for quality control. The first production run is where you find out what your tolerances actually mean in practice versus on a drawing. Having that feedback loop close where you can walk the floor, catch a fit issue before it's in 500 units, and talk directly to the person running the line is worth far more than the cost delta at that stage.
The Shenzhen dynamic you described around shared supply chains is real, and it's structural. The factories aren't doing anything wrong by their logic; they're optimizing for utilization. The misalignment is that their incentive is throughput, and yours is differentiation. US assembly doesn't fully solve that either, but it buys you something Shenzhen can't: the ability to control what actually ships under your name while you're still figuring out what the product needs to be.