Wabi Pivoted Because Users Would Not Build. Would You Notice?

In November 2025, Eugenia Kuyda raised $20 million for an app that worked like YouTube for software. You typed a prompt, Wabi built a mini app, and you shared it with friends.

Eleven months later, she shipped something different.

On October 19, 2026, TechCrunch reported that Wabi had launched for everyone as a personal AI assistant that works like a messaging app. The agent now decides which apps to build for you. The reason for the change fits in one sentence from Kuyda: people want personalized software, but they may not want to spend time building it.

That's a very expensive sentence to learn. And you can learn it cheaper.

Wabi swapped a creator tool for an assistant. The first version, covered by TechCrunch in November 2025, let anyone make apps from prompts and post them for others to like, comment on, and remix. It was pitched as "YouTube for apps."

The 2026 version drops the creator role. You connect your email, and Wabi 2.0 automatically spins up several threads and small apps for you. A health thread might hold a weight-loss tracker. A family thread might hold a local kids' events app. You can still ask for a custom app, but you don't have to.

Same company, same founder. The user's job changed from "maker" to "person who gets things done."

That's a pivot in who does the work, not in what the product can do. Those are the pivots that often come from watching real behavior instead of reading a market report.

Because most people want the result, not the process. Think about photos. Millions of people take pictures, but far fewer edit them, and fewer still post tutorials about editing.

Creator platforms have a famous imbalance. A small slice of users makes the content and a huge slice consumes it. Kuyda's own quote points at this. Her team saw that non-coders liked the idea of personal software but didn't want a new hobby.

Here's the thing: a prompt box feels easy until you stare at it. What do you type? What should the app even do? Blank-box fatigue is real, and it hits hardest with the exact audience Kuyda named in 2025, people who have nothing to do with coding.

If your product asks users to bring the idea, the effort, and the taste, you're betting on a rare kind of customer. Maybe that's fine. But say it out loud before you build for it.

Watch what people do in the first ten minutes, then count who comes back. You don't need a big round to run this test. You need about fifteen real people and a rough prototype.

A few signals that your users don't want the work you're asking for:

• They ask you to "just set it up" for them
• They finish the first task, then never start a second one
• They copy a template instead of making something new
• They describe the result they want, not the steps to get there
• They love the demo you showed and ignore the tool you gave them

Kuyda hasn't said which of these her team saw. Don't assume. But any of them is worth logging, because founders usually hear them as compliments. "This is so cool" often means "I'd like you to do this for me."

Write the observations down within an hour of each session. Memory edits the story fast, and it always edits in your favor.

It keeps what worked and changes the part users avoided. Wabi kept its threads, its custom apps, and its idea that software should feel personal. It changed who triggers the building.

Compare that to a pivot where a team scraps everything and starts over. Sometimes that's right. Often it throws away distribution, learning, and trust.

A useful way to size a pivot is to list your product in three buckets:

1. What users touched again and again
2. What users touched once and left
3. What you built that users never noticed

Keep the first bucket. Rethink the second. Cut the third without mourning it.

Wabi's shift also moved the product toward group use. Kuyda said she doesn't think anyone has "cracked multiplayer" for agentic apps yet, and showed shared threads with her family and her partner. That's a bet, not a result. But it's a bet built on top of something people already do, which is share plans in a chat.

Long enough to test it properly, short enough that it doesn't become an identity. Kuyda's story has a longer clock than most. She presented IO, a conversational concierge for booking restaurants and vacations, at TechCrunch Disrupt in 2014. IO was the first product from her startup Luka, which later became Replika, the AI companion company. She left as Replika's CEO in 2025 and started Wabi.

So the assistant idea she pitched 12 years ago didn't die. It waited. She told TechCrunch that IO "still worked" back then, but it's nowhere near what exists today, because the deep learning advances behind modern language models hadn't happened yet.

Timing is a variable you can't control. Your response to it is the part you can. Some ideas fail because they're wrong, and some fail because the tools weren't ready. The way to tell them apart is to ask what specifically has changed since you last tried. If the answer is a concrete capability, a cost drop, or a behavior shift, the old idea might deserve a second look. If the answer is "I still like it," keep walking.

You don't have to wait for a funded crisis to ask the question Kuyda's team asked. Here's a short list worth putting on paper.

What do users do for themselves in my product, and what do they hope I'll do for them? Where do they stall? Which of my features do I demo most, and which do users open most? If I removed the "make it yourself" step tomorrow, who would complain?

Then map the answers against your plan. You can do that in a spreadsheet, a Notion page, or a planning tool like Foundra that walks first-time founders through each section of a business plan. The tool matters less than the habit of putting your assumptions where you can see them and cross them out.

Add a date next to every assumption. A belief you wrote in March and never checked is just a rumor with your handwriting on it.

Wabi is free for now. The company hasn't chosen a business model, and TechCrunch reported it is in talks to raise more funding. Back in 2025, part of the $20 million pre-seed was earmarked to subsidize usage until a monetization model was settled.

That's a luxury. Most first-time founders don't have angels like Naval Ravikant, Garry Tan, and Justin Kan writing checks, all of whom were named as backers of the pre-seed round.

If you're bootstrapped or running on a small raise, a pivot clock runs differently. Set a spend limit for the test before you start it. Decide in advance: if fewer than X of my fifteen testers return in week two, I change the part they skipped. Write X down.

And keep the old product running if people pay for it. A pivot should add a better path, not cut the one that keeps your lights on.

Users often love the idea of a creative tool and then ask you to do the creating. Watch behavior, not applause. Pivot the part users avoid, keep the part they touch. Date your assumptions. Set test limits before you spend, not after. And remember that an old idea can come back when the tools finally catch up with it.

Kuyda's own timeline runs from a 2014 pitch to a 2026 launch. Yours probably shouldn't take that long. Learning cheaper is the whole game.

What did Wabi launch on October 19, 2026? TechCrunch reported that Wabi launched for all users as a multiplayer personal AI assistant that works like a messaging app, with agentic threads and small apps inside them. It is free to download for now.

How is the new Wabi different from the original? The original, covered in November 2025, let users build and share mini apps from prompts. Wabi 2.0 has the agent decide which apps to build, based on what you connect, such as email.

How much did Wabi raise? Wabi raised a $20 million pre-seed round, announced in November 2025 with angel investors including Naval Ravikant, Garry Tan, and Justin Kan.

How do I know if my users want to build or just get the result? Run short sessions with real people and watch whether they start a second task, copy a template, or ask you to set things up for them. Those behaviors tell you more than survey answers.

Is it too late to pivot if I've already raised money? No. Wabi pivoted after raising a large pre-seed round. What matters is whether you can name what users avoided and what you will change in response.

Should every pivot change who does the work? No. Some pivots change the customer, the price, or the channel. The useful habit is to ask what users skipped, then decide which part of the product that points to.

1. Replika's founder built the AI assistant she pitched at Disrupt 12 years ago, TechCrunch, October 19, 2026
2. Replika founder raises $20M pre-seed for Wabi, the 'YouTube of apps', TechCrunch, November 5, 2025
3. Replika founder raises $20M pre-seed for Wabi, the 'YouTube of apps', Yahoo Finance
4. Wabi raises $20M, Founders' Daily Briefing
5. Wabi opens its personal AI assistant on iOS, Android and Mac, Unite.AI