What Notion, Figma and Pokemon got right about multi-player

A retrospective on building (and shutting down) Undesk and Tabl

When a startup ends, it leaves behind a strange, quiet feeling. There’s regret about what didn’t work, relief that the uncertainty is finally over, and a lot of unanswered questions.

It’s also one of the rare opportunities in your life when you’re forced to slow down and reflect on what actually happened.

This post is a retrospective on building my first startup, Undesk and Tabl, and the lessons I learned along the way. I’m writing it in the hope that it might resonate with others who are navigating a similar path.

After making the personal computer mainstream with the Mac, Steve Jobs started talking about a new idea: interpersonal computing.

Three decades later, in 2020, it felt like the right time to think about this idea again. COVID pushed everyone into remote work, and when things settled, people started having two opposite opinions: RTO vs WFH. Search “remote work” on X or Reddit and you will see very strong (or extreme) opinions on both sides.

But one thing was clear. Remote work still had a lot of friction. Friction so real that companies were willing to tolerate half-empty offices, hours of commuting, and even losing talent unwilling to commute.

So we set out to remove that.

Undesk — A meta-layer virtual office

Before Undesk, I had spent years working in VR startups as a PM. I was one of those people who truly believed headsets would replace desktop computers soon. But it took way longer than expected. A decade later, almost nobody uses a headset on a daily basis today.

But even back then, it was clear that virtual avatars could create a surprisingly strong feeling of human presence. It felt powerful enough to replicate the energy of a small startup office.

Headsets weren’t ready for everyday single-player computing. They were heavy, uncomfortable, and simply not something people could use for more than 30 minutes. But VR technology itself could enable multi-player computing. So our first prototype mimicked a headset experience, but on desktop.

Another thing I believed was that a virtual office had to be a meta-layer app, not a peer-layer app.

Peer-layer apps take your attention one at a time. When you open Slack, you move out of Figma.

Meta-layer apps work differently. They quietly live alongside other apps. Dropbox syncs files in the background. 1Password suggests passwords right when you need them.

A real office feels more like that. It’s just there, in the background of your work. That meant we had to nail the switching experience, making the transition between focus and collaboration feel blended and effortless.

What worked

Using 3D was controversial considering CPU resources, but it worked. Even tiny animations like breathing and typing made human presence strong.

The launcher wasn’t for everyone. Some people couldn’t stand having even a small window floating on the desktop. But for many early users, it quickly became part of their daily workflow. When your “office” is buried among 50 open tabs, you stop going back to it. The launcher fixed that, and the keyboard shortcut (used to toggle the window open or closed) quickly became one of our most-used interactions.

What didn’t work

It was easy to create an initial wow moment, especially when people opened up the launcher and saw the 3D office for the first time. But it’s still hard to make the product sticky. Many internal champions helped us spread the word, but struggled to drive long-term adoption.

What we learned

A virtual office needs a reason to exist even when you’re alone. Users need to get value first just by being there. Communication only comes after.

Undesk 2.0 — A personal productivity tracker

Magic App Sharing was one of the most popular features among early users. It shows which apps you are actively using, both native and web, and shares that activity with your team in real time.

At the same time, first impressions were often negative. Many people felt the feature was too invasive. Still, around 90% of users kept it turned on. When they tried turning it off, they realized how much context and awareness the app status had been providing.

Since we were already detecting app activity, we wondered: People track weight, calories, spending, and sleep. Why not track how they work?

Quantifying productivity was not straightforward. We first categorized apps between Deep Work (coding, design, analysis) vs Shallow Work (chat, email, meetings), based on Cal Newport’s Deep Work theory. We then measured how active users are on Deep Work apps without frequent context switching. The goal was to infer flow state patterns automatically, without manual inputs.

This was meant to be Undesk’s single-player mode.

A virtual office is a multi-player product. It has no value if you’re the only one inside. But if we could give people a compelling solo experience, we could build a new game cycle with stronger retention: something like training your first Pokémon before battling with your friends.

What worked

Retention improved dramatically. Once the tracker became a habit, people came back to their virtual office every day. Over time, complaints about app detection faded. Many users even started adding their own app stacks to improve accuracy.

What didn’t work

The core users of a productivity tracker (individual contributors) and the core users of a virtual office (managers) were fundamentally different. People loved their own dashboard but rarely invited teammates. The game cycle we tried to create didn’t really complete.

What we learned

Although we tried so hard to position ourselves as a meta-layer, a virtual office was still an extra peer-layer communication app. It was too hard to convince people to add another app stack in addition to Slack and Zoom. We needed to take a higher risk to be a true meta layer, in a place people already spent most of their time. We had to sit one layer above.

Tabl — A multi-player web browser

In summer 2024, Techstars gave us a conditional offer: “We like your team, but not the idea. If you’re willing to pivot, we’re willing to fund you.”

By then, we already sensed the virtual office window was closing. The category peaked in 2020, then quickly lost trust. We wanted to revive it, maybe even redefine it, the way Zoom redefined video calls when Skype couldn’t. But the market sentiment was negative, and the timing felt wrong.

When we got this offer, we were shocked. At the same time, we felt strangely relieved and excited. It felt like we were given one more chance.

So we went back to the whiteboard.

We stepped back and looked at where people actually spend their time when they work. One insight kept resurfacing from user interviews: people spend most of their working time in a browser. If you check your desktop screen time, chances are the browser is the most used app every day.

The browser is where problems are solved, documents are written, code is reviewed, and conversations happen. The browser is where work happens.

This could be a meta-layer. This could be the engine of team momentum. What if we could make the browser multi-player?

More on this blog post, but the core idea is to bring what’s happening in workplaces in a browser.

Productivity layer
This is where the work happens, solving, designing, building, and shipping. The ultimate goal of teamwork is to maximize output from this layer. But to do that, the next two layers must function well.

Collaboration layer
This exists to unblock the productivity layer. It includes activities like bug reporting, white-boarding sessions, and company-wide strategy meetings. More collaboration isn’t always better, but insufficient collaboration is a direct drag on productivity. The moment you delay a decision until tomorrow’s standup, your team starts to slow down. Delayed feedback = delayed momentum.

Presence layer
This layer lowers the friction of collaboration and amplifies productivity. During the pandemic, we thought we could get by with just the first two. But now we’re seeing why presence matters — one big reason why major companies are pushing return-to-office.

Tabl attempted to combine all three into a single window.

You can see what your teammates are working on real-time. Landing on the same tab instantly created shared context. You can bring your team into your tab without switching apps.

After 100 user calls and 12 intense weeks at Techstars, this is where we landed.

After Demo Day, we spent another 8 months building a browser from scratch. We knew it wouldn’t be easy, but we decided to go all in.

What worked

Our positioning resonated strongly, and Tabl hit #1 Product of the Day on Product Hunt. It was the moment The Browser Company announced to stop building Arc. The timing was great to attract early adopter community looking for a new productive browser to experiment.

What didn’t work

The browser is, as mentioned earlier, where you spend most of your time. If one tiny thing doesn’t work, you immediately notice it. No tab groups, no touchpad gestures, extensions misbehaving, the list goes on. There are thousands of table-stake features that Chrome does perfectly. We tried to close this gap feature-by-feature, but we could not meet the expectation in 8 months.

But product quality was only part of the problem. Adoption brought a different challenge.

Moving from Chrome is way easier than moving from Slack or Notion. Bookmarks and passwords are very portable compared to chat messages or block-based database. But convincing an entire team to switch browsers turned out to be a different game. If the first person doesn’t bring the second person, the multi-player loop dies instantly.

What we learned

First, we misunderstood the single-player mode again.

With Tabl, we assumed that web browsing itself could serve as the single-player experience. But we were wrong. A single-player mode has to offer unique, personal value before any team features kick in.

You switch to Notion from Evernote because of its customizability. You switch to Figma from XD because of its UX. You start using it alone anyway. When you invite your team, the value multiplies, but you don’t have to. The initial aha moment has to come before a user invites their team.

Second, when evaluating a product like a browser, people focus more on deducting points for flaws, less on adding points for strengths.

Rather than “reasons to switch,” they go through a checklist of “reasons not to switch.” This is probably true for established product categories. Most users of Windows, Chrome, Slack today keep using them not because they love them. Because other new alternatives cannot do certain little things which you can do now with your current product.

I personally noticed this when I switched from Google Pixel to Nothing. Attracted by its unique industrial design, my wife switched to Nothing as soon as it came out. But we noticed that it had one critical bug in Android Auto, which I have to use every day in my car. So I waited and stayed with Pixel, the most reliable Android phone which I don’t really love. When the bug finally got fixed, I was so happy to be able to switch.

A big reason to switch (the bold industrial design) was not enough. Removing reasons not to switch was the trigger. And I can see that, for Nothing, it’s just one of millions in the list. In established product categories, a single critical flaw is often enough to keep people anchored to what they already know.

What happened

This time, there was a sense of momentum we hadn’t felt with Undesk. After the Product Hunt launch, we built a solid top of the funnel, and for the first time, we started to feel that quiet sense of “maybe this is it.” But the viral loop never formed:

• 56% of users churned before inviting anyone
• 89% of invited users invited only one person

Without multiplayer value, there was no strong reason to switch browsers. Eventually, we ran out of cash.

What I’d do differently

Strategically, we should have started with a true single-player mode. Even if the end goal was multi-player computing, the first question should have been:

1. Why should I switch today?
2. Why should I ask my team to switch tomorrow?

You can’t skip the first to reach the second.

In fact, we had single-player ideas from the very beginning. But our attention was consistently pulled toward multi-player UX. That bias largely came from our roots in Undesk, a virtual office, where collaboration was always the starting point. The order should have been reversed.

Multi-player products don’t fail because collaboration is hard. They fail because the single-player loop isn’t compelling enough to earn the right to collaboration.

Operationally, we needed to ship faster. It may sound like a cliche, but looking back, there were things we over-engineered and optimized too early. If we had learned these lessons even 3 months earlier, we might have tested the single-player ideas sooner, gained traction, raised another round, and expanded from there. Startups succeed from a chain of “ifs.” And we missed a few critical ones mostly because of lack of speed.

Personally, I should have taken my mental state more seriously. This was the first time I experienced real burnout. Toward the end of the runway, I wasn’t performing at the level I expected from myself. Sleep, food, exercise, none of it can be neglected. Burnout didn’t just reduce my output. It reduced my ability to tell when something wasn’t working. For me, this was the biggest hard earned lesson.

What’s next

We didn’t make it. But I still believe multi-player computing is inevitable, for two reasons.

1. The new browser war has begun.

OpenAI Atlas, Perplexity Comet, Dia by The Browser Company, and Chrome with Gemini. After a decade of Chrome domination, the idea of switching browsers finally feels realistic again. The platform is in flux, and that creates opportunity. This might be the only real chance to build at the browser layer.

2. Human-to-AI communication is still single-player.

People spend less time with human-to-human communication and more time with human-to-AI communication today. AI took over not just Google time but also Slack time. But soon, AI will blend into human-to-human communication, becoming part of multi-player interactions.

So I really hope someone eventually cracks it.

As for me: I’m still figuring out what I’ll build next yet. But I know I’ll keep building. It feels like one of those rare windows where the timing is right to build something truly game-changing.

If any part of this story resonated, let’s chat.

I’ve been in heads-down building mode for a long time, and I want to spend more time talking to people again. Whether we’ve worked together before or haven’t met yet, I’d love to hear what you’re working on, what you believe about the future of work, or what you think is still broken.

And if it turns into building something together, that’s even better.

Before closing, I want to say thank you.

To our investors, thank you for trusting us through uncertainty and backing ideas that didn’t come with guarantees.

To my co-founder Sho, thank you for building alongside me through the late-night calls, the hard conversations, and the moments when it felt impossible.

And to everyone who tried our products and shared feedback, thank you for giving this journey meaning.