How Wispr Flow found PMF through a pivot

Want to read more write-ups like this? Subscribe to the Product Hunt weekly newsletter for a new one every Sunday 🌟

My childhood dream was to build a real-life JARVIS, Tony Stark’s AI assistant. And in 2021, it looked like I’d have my shot.

It was in between the launches of GPT3 and ChatGPT. We were starting to see things happening with large language models, and I believed that in the next few years, we'd all be talking to our devices. But if everybody was talking to devices, there’d be no privacy and we’d be disturbing other people.

I thought: If we could build a device that lets you communicate with everything around you silently, that would be a game-changer.

So we spent half a year building an early version of that. It was a small wearable device that converted neural signals from silent speech into text or voice. Once we had it working in a limited capacity, we raised our first round of funding and were off to the races.

Over the next three years, we built an incredible team, with some of the best PhDs in neuroscience, machine learning, and signal processing. We used our funding to build a 40-person company, and we actually got this crazy idea to work. It was a pair of headphones that you put on, and when you silently speak, it would convert your brain signals into a voice that sounds exactly like you.

It literally felt like magic.

The Hard Truth About Hardware

But that was the problem.

As soon as we had it working, we hooked it up to ChatGPT, Siri, and Alexa. The results weren’t great. Because the stuff in your mind is extremely rambling.

Okay, we’ve got this, I thought. We dove into creating something that could structure the rambling in your head into something you could actually send. We called this Project Flow. We were essentially building this as the operating system for the hardware device.

But that didn’t fix bigger issues with the hardware — there was no consumer market for it. Around this time, I saw the Humane AI Pin launch and crash. I saw rabbit’s pocket companion launch and crash. I saw other hardware devices launch…and crash. Back then, consumer voice workflows hadn’t crossed the chasm. Our hardware product would have flopped — even though it would have been the most magical thing in the world.

The Conversation That Changed Everything

After a board meeting in mid-2024, my co-founder Sahaj and I started talking about whether the hardware device was really the right path forward. At first, I pushed back. This was my childhood dream, something I’d been working toward for the last 16 years. But over the next few weeks, we kept coming back to the same conclusion: As much as we loved what we’d built, the best move for the company was to let it go.

The software that existed at the time simply wasn’t reliable enough. Moreover, none of it was very fluid — it didn’t work across apps. Bottom line: Voice dictation just wasn’t good enough to get people to adopt it as a habit and replace their keyboard as their default input. It didn’t matter how good the hardware was if the software wasn’t ready.

Luckily, we realized we had already been building two companies in parallel: one for the Wispr hardware and one for Flow software. But we were still a small startup. We knew we could only get one thing to PMF at one time, and we just had to pick our battles.

We saw that if we built the software layer first — a magical voice experience that actually works and can replace your keyboard — we or someone else would have the opportunity down the line to build a hardware device that lets people do it even better.

Finally, after a few weeks of hard discussions, I concluded the Wispr hardware device would not have worked as a consumer product.

Now, we did have options to stay in the hardware business. Had we gone B2B, made it a healthcare product, maybe a defense product, we could have gotten a lot of funding. Those paths might have been fundable, but they lacked founder-market fit. Life is short. I believe you should do the things that excite you and get you up in the morning. And those options wouldn’t have gotten me out of bed.

Hardware –> Software

By July 18, 2024, we stopped working on the hardware device completely and gave all our attention to Wispr Flow, the voice dictation platform.

Our investors supported us. They had seen some inklings of this — and seen plenty of founders unable to make the tough call of killing their baby.

Still, that difficult situation forced some other difficult decisions. We initially thought we would keep a handful of people on the hardware team so that we could keep doing some experiments there. We had amazing people, including one of the best systems neuroscientists in the world. But we chose to do a clean cut. I didn't want to do a second round of layoffs later. And I wanted people to feel certainty. So we laid off almost everybody. In late July, we went from 40 employees to 5 employees overnight.

We knew it was the right thing to do, but that didn’t make it feel any better. And even with all the heartache, we didn’t even know whether it would be successful.

Because we didn't know if this was a thing that was going to work and because I was too shellshocked from laying off our team, I was very wary of hiring for a few months. I didn’t want to be in a situation again where I’d have to let people go. So even until about January this year, we were still under 10 people.

Signs of PMF

We were initially planning to launch Wispr Flow in January. That would give us four months to get it ready.

But I expedited the timeline. Six weeks, gun to your head: What would you do to launch this product? In that month and a half, we spun up an entire marketing operation and did a huge drop on October 1, 2024.

It was actually pretty successful. We got millions of views and hit #1 on Product Hunt for the day and the week. Two months before that, we weren't even a consumer software company.

By January, we were seeing long posts on LinkedIn and X about how the product had changed people’s lives. We also saw ~20% of our users convert to paid (vs. the usual 3–4%), and organic growth of ~90% month-over-month in January and February. On average, users were doing around 100 dictations a day and typing only 25–30% of their total input on a keyboard.

Now we knew we had product-market fit, strong user sentiment, and great word of mouth. That’s when I finally felt confident enough to start hiring again.

I would never have started a voice dictation app company. It doesn't sound sexy, doesn't sound world-changing. But when we talk to users, they say this has had a bigger impact on their life than literally anything else. That was the only signal we really needed.

Four Takeaways

It took a lot of hard calls to get to that point. Here are the four that mattered most.

#1: Pivot fast to get out of the in-between zone.

In a week, I went from “I don't think I think we should pivot” to “Let’s execute.” Our timeline for that transition to launch was six weeks, which I hear is much faster than average. But you have to move fast to get out of that grey area. Because otherwise you're paralyzed, which will make you a terrible leader. That fast timeline also gave people a goal to drive toward, which was really helpful in that transition period.

#2: Cut once and cut deep.

If you’re unsure about whether to keep people, you'll be doing everybody a favor by letting them go. You don’t want to retain people you're not sure you have a role for. They’ve just seen layoffs happen, and they want certainty. Everybody on the team who stays should feel like you're a “hell yes” on them.

#3: You will feel like shit.

Your team still needs you to show up. I had to let go of all of my favorite people, the ones who had signed up to help me build my dream. The people who remained were as shellshocked as I was; we had a 22,000 square foot office with five people in it. I didn't know what the company was doing, and all I wanted to do was hide. But I couldn't do that, because the team needed me to be strong. So I showed up every single day, first person in and last person out, and tried to give everybody a deep feeling of certainty. Being there with the team in the trenches helped get morale back up.

#4: Get your facts from your customers and keep your opinions to yourself.

We spent three years building this device that sounds magical. It sounds really sexy. But there was no customer who saw that and said, “I know exactly what I'm going to do with it.” We just thought it would be like Field of Dreams — if we build it, they will come. Sure, that happens sometimes, but we had no signal to suggest that. When someone tells you what they won’t buy, believe them.

More Than a Pivot

Looking back, the pivot wasn’t just about changing products — it was about changing how we made decisions. We stopped chasing the sci-fi dream and started building what people truly needed. That shift gave us our second act, and it’s the reason Wispr Flow exists today.