Chapter 3: The Product That Almost Wasn't
There's a shift that happened in the technology world over the last few decades that most people lived through without fully noticing.
It started in garages. Literally. Apple in Los Altos. HP in Palo Alto. The story of American technology innovation was, for a long time, a story about people building physical things in small spaces with limited resources and enormous ambition.
From Garages to Shenzhen
The garage was the symbol because it was real: this is where hardware was born, where you breadboarded circuits and argued about industrial design and figured out, by hand, what the product should be. Then China opened up. The Pearl River Delta became the most sophisticated manufacturing ecosystem the world had ever seen. Specialized assembly lines, workers with decades of accumulated craft knowledge, supply chains so dense and interconnected that you could source almost any component within an hour's drive of Shenzhen. What had taken months in American factories could happen in weeks here. What had required enormous capital could be done at a fraction of the cost.
The result was a quiet but total reorganization of where things got made. As NPR reported even in the early days of this shift, many of the cool tools consumers craved were dreamed up in the United States, but when it came time to turn those ideas into real products, the manufacturing work was sent to China. "Designed in California, assembled in China" became the template, and for a long time it worked well for everyone involved.
But then software ate the world, and the valley moved on entirely. MIT Technology Review captured it well: the Pearl River Delta's mix of factories, component suppliers, service providers, and skilled workforce had become something genuinely difficult to replicate anywhere else. While Shenzhen quietly became the hardware capital of the world, Silicon Valley pivoted its identity entirely toward software. The status and the capital followed.
A software company with enterprise potential could raise at multiples that a hardware company could rarely touch. The logic was scalability: software ships to a million users overnight, hardware has to be manufactured and packaged and shipped and returned and replaced. The margin profiles were different. The growth curves were different. The stories investors told themselves were different. Data from over 400 M&A transactions between 2015 and early 2025 shows software companies achieving a median valuation of 3.0x revenue compared to just 1.4x for hardware, a gap that reflects the scalability and recurring revenue advantages investors associate with software.
The Software-Only Temptation
Hardware became, in the minds of a certain kind of venture investor, the harder bet. Not because it wasn't valuable, but because it didn't scale the way a SaaS dashboard scaled. As Razer's CEO wrote in TechCrunch, venture capital and tech talent had been fawning over software for years, while consumer electronics makers attracted a fraction of the investment, in part because, as Marc Andreessen famously put it, hardware is hard. The question that got asked in every hardware pitch meeting, usually early, was always some version of: but what's the software story? Where's the recurring revenue? How does this become enterprise?
I'd spent my career in that world. I understood the math. I'd watched Anker navigate it, building a hardware company that was genuinely global, genuinely profitable, genuinely respected, while the valley talked about software multiples. It was possible. It just required a different kind of discipline. So when Memoket became real enough to have a strategic question, the question wasn't surprising. Do we build hardware at all?
We had the algorithm. We had the prototype app. We had a problem we knew was real and a solution that worked. The software path was cleaner in almost every way that a spreadsheet could measure. Lower capital requirements, faster iteration, easier to scale, better valuation optics for the fundraising conversations we knew were coming. The argument for staying software-only was not a bad argument. I want to be clear about that. We had it seriously, more than once.
But we kept coming back to the same problem. Recording on a phone sounds obvious until you actually try to use it as your primary capture device. The phone does too many things. A call comes in and the recording stops. A notification pulls your attention. The battery that was at 40 percent when the meeting started is at 12 percent an hour later and you're making decisions about whether to let it die. The phone is the most contested piece of real estate in your life. Asking it to hold your context while also doing everything else it does is asking too much of it.
The Apple Watch was better in some ways. Smaller, less intrusive, already on your wrist. But the same interruption problem existed, and the storage and microphone quality weren't built for what we needed. I'd already seen, in my consulting period, what happened when you put a dedicated recording device on the table. The room changed. People felt watched. The conversation became less honest, less useful, exactly at the moment you were trying to capture it.
Adding to a Wearable
So we had a phone that was too busy, a watch that wasn't quite right, and a dedicated device that worked technically but failed socially. And we had a team with decades of hardware experience sitting around a table in Shenzhen. Huilong said it first, I think, though Elisa and I were already thinking it. We know how to build things. We know this ecosystem. We know these suppliers. If the existing hardware solutions are all wrong in the same way, the answer isn't to accept the best available option. The answer is to build the right one. The question then became: what does the right one look like?
We looked at what else was in the market. AI wearables were having a moment. Pins you wore on your jacket. Earbuds with always-on recording. Devices that hung around your neck. All of them shared the same basic assumption: the AI wearable was a new category, a new thing you'd add to your life, a new device competing for space alongside everything else you already carried. We didn't think that was right.
The people we were building for, founders, SMB owners, consultants, sales professionals, already wore things. They had Apple Watches or Garmins or Whoops or Oura rings. They'd already made decisions about what belonged on their body. Asking them to add another device to that inventory, to charge another thing, to remember another thing, to explain another thing to the person sitting across from them, was a real cost. Even if the device was small, the cognitive overhead wasn't.
We looked at our wrists. All three of us wore Apple Watches. What if instead of adding a wearable, we added to a wearable? The idea sounds simple now. At the time it felt like it opened something up. Not a new device competing for your attention. A layer that attached to the device you already trusted, already charged, already wore every day without thinking about it. Something that disappeared into your existing life rather than asking your existing life to make room for it.
Privacy as a Design Constraint
We reached out to two former Anker colleagues, both specialists in industrial design. We told them what we were trying to do. Not a spec, not a brief, just the idea: small, light, premium, attaches to the bottom of an Apple Watch band, invisible until you need it. Within weeks, they had a 3D print.
I remember holding it for the first time and feeling the thing that you feel when a physical object is right. It's hard to describe to people who haven't spent time in hardware. There's a quality to a well-considered object that you register before you've consciously analyzed it. Its weight distribution. The way it sits in your hand. The sense that someone thought carefully about every surface. It needed adjustments. Of course it did. First prototypes always do. But the direction was right. You could feel that too.
It attached to the bottom of the watch band cleanly, sitting just below the Apple Watch face, co-wearing rather than competing. It looked like it belonged there. Not an afterthought bolted onto something else. A considered addition to something people already loved. It was light. Light enough to forget you were wearing it. Small enough that it didn't change the profile of the watch. But with dual microphones and enough on-device storage to hold 400 hours of audio, and a 20-hour battery that outlasted any meeting you were likely to have.
We'd thought carefully about privacy from the beginning. Not as a compliance checkbox but as a genuine design constraint. The device we were building was powerful precisely because it captured real conversation, the unguarded kind, the honest kind. That power came with responsibility. The always-on approach, the one some AI wearables had taken, wasn't right for us. We didn't want a device that was recording without the user choosing to record. Press to start. Press to stop. That was the interaction model from the beginning.
But we also wanted the people in the room to know. Not because we were required to tell them, though in many jurisdictions you are, but because we believed it was the right way to build something people could actually trust. A device that operated in secret, even legally, would change how people felt about it once they found out. We'd already seen what happened when a recording device sat visibly on a table. We didn't want to solve that problem by hiding the device. We wanted to solve it differently.
The Product Built Right
The red LED came from that conversation. When Gem is recording, a small red light is visible on the device. Not hidden, not subtle. Visible to everyone in the room. The same red light people have understood for decades on cameras, on phones, on anything that's capturing. It signals: this is on. You know. We all know. What we found, when we started testing it, was that the red LED didn't create the awkwardness we'd worried about. People were more familiar with recording indicators than we'd assumed.
What made the note-taking device attached to my iPhone feel intrusive wasn't the fact of recording. It was the form factor: a rectangular slab sitting on a conference table like a surveillance device. The Gem, on your wrist, with a small red indicator, read differently. It read like a tool. A professional one. Something you'd chosen to use rather than something that had been placed in the room. The awkwardness didn't disappear entirely. It never does. But it was manageable in a way the earlier device never was.
We had the design. We had the algorithm. We had the app. We had a manufacturing partner lined up for early 2026, someone who could take us from prototype to samples to production run. And then that partnership fell apart. I won't go into the details because some of them are still being untangled and none of them are particularly interesting. Logistics. Misaligned expectations. The kind of supplier problem that happens in hardware development more often than anyone who hasn't done it would believe. You plan for it in the abstract and it still surprises you when it arrives.
What I will say is that it cost us time we didn't have a lot of and required us to restart a process we thought we'd already completed. And then we found a new partner. I've been in this industry long enough to know that manufacturing relationships are not transactional. The best ones are collaborations. Your manufacturing partner sees things you don't see, knows things about materials and tolerances and production realities that no spec document fully captures.
When you find a partner who is genuinely curious about what you're building, who pushes back when something won't work and offers alternatives rather than just rejections, who treats your product as something worth caring about, that's not a vendor. That's a co-builder. The new partner was that. More flexible, more engaged, more willing to work through problems in real time rather than at arm's length. What felt like a setback turned into a structural upgrade. The product that almost wasn't became the product that was built right.
Memoket Gem captures your conversations so you can stay present in them. See how
