Wabi 2.0 Turns Your AI Into a Shared Group Chat

Update, Oct. 20: launch-day confirms the 'proactive group planner' pitch

A day after launch, the coverage is settling on the same frame Kuyda pitched: Wabi announced the release in an official X launch thread, describing the assistant as 'highly proactive' — one that notices what would help a group, makes plans, keeps things moving, handles routine tasks, and builds small apps when they are useful (Unite.AI, Oct 19). The company's download page offers direct App Store, Google Play, and Mac links with no waitlist step, and the App Store listing from Wabi, Inc. describes it — free for now — as 'an assistant for getting things done with other people' (Unite.AI). That is the bet in one line: not a smarter chatbot, but a group planner with a toolbox. Town AI's $1B valuation shows investors are still paying up for agent startups while Wabi itself says it is in talks to raise more funds.

Twelve years ago, Eugenia Kuyda walked onto a TechCrunch Disrupt stage and demoed a chatbot that could book your dinner and plan your holiday. It was 2014 — years before the phrase "large language model" meant anything to anyone — and the demo, in her own words, was "pretty magical." On Monday, she finally shipped the grown-up version.

There's one detail she couldn't have promised back in 2014. This time, your AI assistant arrives with a group chat attached — and your partner can ask it, in plain English, whether you're free next Thursday and whether you bought the tickets. Welcome to Wabi 2.0, the "multiplayer" personal AI assistant that Kuyda's startup launched to everyone on Monday. We checked the timeline ourselves: her IO pitch dates to October 2014, so the "twelve years" lands almost to the week.

Twelve years from IO to Wabi

The backstory matters here, because almost nobody in AI has been at this longer than Kuyda. IO — a lifestyle-concierge bot from her first startup, Luka — predated the deep-learning breakthroughs that made today's chatbots possible. "It still worked," Kuyda told TechCrunch in an exclusive interview this week. "It was pretty magical ... but, of course, it's nowhere near what we have today."

Luka eventually became Replika, the AI companion app that turned Kuyda into one of the best-known founders in consumer AI. She left as Replika's CEO last year and founded Wabi, which announced a $20 million pre-seed round last November (reportedly led by Andreessen Horowitz). The first version of Wabi was a social platform where you typed prompts to create shareable mini apps — a sort of YouTube for tiny software. Then, on September 28, the company posted on X that it was becoming something else entirely: "a new kind of messenger," invite-only, described by Kuyda as "a personal agent that does stuff for you and builds the interface you need in the moment."

Panel on stage at TechCrunch Disrupt Berlin 2018
The TechCrunch Disrupt stage, Berlin 2018 — Eugenia Kuyda first pitched her AI-assistant vision on a Disrupt stage in 2014 (illustrative photo). Photo: TechCrunch, via Wikimedia Commons (CC BY 2.0).

After three weeks of invite-only testing, Wabi opened to everyone on October 19 — free downloads for iOS, Android and Mac, no waitlist, per Unite.AI's launch-day report. The company still hasn't settled on a business model, and Kuyda says Wabi is already in talks to raise more money. Free now, priced later, funded by venture capital: the classic consumer-AI opening gambit.

An assistant that builds its own apps

Strip away the marketing and Wabi 2.0 is two ideas bolted together. The first: a messenger-style app organised into multiple agentic threads, each dedicated to a task or a long-term goal, instead of one endless scrolling conversation. The second: inside each thread, the agent builds and runs custom apps — a weight-loss tracker living in your health thread, a finder for nearby children's events inside the family thread.

The twist is who does the building. Last year's Wabi asked you to dream up the mini apps. Over the past year, Kuyda says the team learned something humbling: people love personal software rather than "one size fits all solutions in the App Store," but they "maybe don't want to spend a lot of time, or any time, building it." So in Wabi 2.0, the agent decides what to build. Connect your email and it immediately spins up a starter set of threads and apps on its own; you can still request custom apps or open new threads by hand. It's the difference between being handed a toolbox and being handed a finished kitchen.

That's also where Kuyda draws the battle lines. "It's doing what Instinct and Muse do, but it also has all these apps, which helps users do more stuff," she said — naming Meta's Muse and the rival assistant Instinct as the incumbents she's hunting. If you want the primer on what today's agentic AI can actually do versus what the marketing claims, start there; Wabi's pitch is that apps, not chat, are where the doing happens.

The multiplayer bet: your AI joins the group chat

The second idea is the bolder one, and it's the reason this launch is worth your attention. Every Wabi thread can be shared — it doubles as a group chat where couples, families or friends plan together with the assistant sitting in the middle. Shared calendars, shared to-dos, shared context. "Today, all the agentic apps are just [conversations] between you and the agent," Kuyda said, "and I feel like it should be a lot more about bringing people together in a group chat together. Not you doing it separately with your AI, and I'm doing it separately with my AI ... I don't think anyone has yet cracked multiplayer."

She's eating her own product, and the walkthrough she gave TechCrunch is the most human part of the story. Her thread with her boyfriend has a row of apps along the top: a live music tracker, an app that asks the couple a question about their relationship every single day, a running restaurant list, shared calendar events, and a list of books they're reading together with suggestions for what to read next. Her family thread tracks her daughter's schedule, finds kid-friendly events nearby, and — this is the detail that stuck with me — quizzes her daughter on Japanese.

The technical argument for splitting conversations this way is genuinely strong. "It's not just everything in one single thread like with Instinct and Muse, where you get buried in the information that it sends you every day," Kuyda said. "Single thread experiences, if you think about it, get worse over time because it's just so much context in one endless conversation." She's right about the failure mode: AI researchers have found that the longer a chatbot conversation runs, the more likely it is to hallucinate, lose the thread, or slip past its own safety guardrails. Wabi's answer is architectural — many small contexts instead of one rotting one.

Inside any shared chat, each person decides how much to expose: email, calendar, even bank account details. "So my partner can ask Wabi in the chat, 'Is Eugenia free next Thursday?' or 'Did she buy those tickets?'" Kuyda said. Access is scoped to individual chats, and — a nice touch — everyone in the chat can see when someone's personal information gets requested. No silent snooping, at least in theory.

Wabi vs Instinct vs Muse: the matchup that decides consumer AI

Kuyda is picking the fight on purpose. "It's doing what Instinct and Muse do, but it also has all these apps, which helps users do more stuff," she told TechCrunch — naming Meta's Muse and the rival assistant Instinct as the incumbents she's hunting. Her case against them is architectural: "single thread experiences ... get worse over time because it's just so much context in one endless conversation" (Kuyda, TechCrunch). The failure mode she names is documented: AI researchers have found that the longer a single-thread chatbot conversation runs, the more likely it is to hallucinate, lose context, or slip guardrails (TechCrunch).

The money, for now, is on the other side. Instinct's $1B raise at a $10B valuation — Instinct, the rival Kuyda keeps naming, is the best-funded of the bunch after that round. Wabi's counter is that apps, not chat, are where the doing happens.

Our take: a real category hiding inside a colossal privacy bet

Separate the two ideas, because they deserve different verdicts. The multi-thread design? Just correct. Every AI assistant today is fighting the same enemy — context rot — and Wabi's answer of many small, purpose-built threads is the first consumer design that treats it as an architectural problem rather than a cosmetic one. Expect it to be copied within a year.

The multiplayer part is where it gets genuinely interesting — and uncomfortable. As a growth strategy it's brilliant: a group chat is a distribution machine, because every thread needs at least two humans. WhatsApp didn't grow on engineering alone; it grew because your friends dragged you in. Wabi is betting the same loop works for AI.

But look at the bargain. A free app, no business model, running entirely in the cloud, asking for your email on day one — while normalising bank account details as shareable objects inside a family chat. Kuyda's privacy answers are real but thin: SOC 2 means the security controls were independently audited, not that your data won't train a model. "Most" user data stays in the U.S. — not "all." You choose which connections to authorise and can disconnect them, per the app-store descriptions, but the product only works if you hand over the keys first.

My stance: multiplayer AI is a real category, not a gimmick — shared planning is one of the few things a solo chatbot is structurally bad at. "My previous company Replika worked on AI that will make people feel ... less lonely," Kuyda said. "To a degree, that is [in Wabi] as well." But the category will be decided by trust, not features. Try it by all means — just keep your bank details out of the family chat until Wabi's privacy story is as ambitious as its product story.

Wabi 2.0 FAQ: your questions, answered

Is Wabi free? Yes, for now — the company openly admits it hasn't decided how it will make money. Expect that to change.

Which platforms does Wabi support? iOS, Android and Mac at launch, with direct downloads and no waitlist.

How is Wabi different from Replika? Replika was an AI companion built to ease loneliness. Wabi is a practical assistant for getting things done, built around shared, multi-person threads.

How does Wabi protect my data in shared chats? Wabi is SOC 2 compliant — its security controls were independently audited, which is not a promise your data won't train a model. The company says "most" user data stays in the U.S. — not "all." Access is scoped to individual chats, and everyone in a chat can see when someone's personal information gets requested. No silent snooping, at least in theory.

What apps can Wabi's agent build inside a thread? At launch, the examples are a weight-loss tracker living in a health thread and a finder for nearby children's events inside a family thread. Kuyda's own threads show the range: with her boyfriend, a live-music tracker, a daily relationship question, a restaurant list, a shared calendar and a joint reading list; in her family thread, her daughter's schedule tracking, kid-friendly event finds, and quizzes on Japanese.

What happened to the companion-AI boom Kuyda left behind? The category is brutal on economics. The cautionary tale is Character.AI's bankruptcy — where Disney bid just $20M for its tech.

Sources

TechCrunch (Oct 19, 2026 — exclusive Kuyda interview; Sep 29, 2026 — Wabi 2.0 invite-only launch); Unite.AI (Oct 19, 2026 — platforms, app-store data practices); Light News 24; RuntimeWire; SuperpowerDaily. All direct quotes are from Eugenia Kuyda, founder of Wabi and former CEO of Replika, as reported by TechCrunch.

Related

AI Agents Explained: What They Actually Do (and What They Don't) — the primer on agentic AI behind assistants like Wabi.

Claude Steers Wealthy Flyers to Tickets It Chose Itself — another AI agent making its own decisions about your bookings.

ChatGPT Drove a Toyota to In-N-Out at 09:41 — when an AI agent took a real-world errand into its own hands.

Daniel Rao covers AI infrastructure, cybersecurity and the business of the frontier labs for TechPulse USA.

Comments

Popular posts from this blog

Nous Research Raises $90M With Trump Jr. Firm on Cap Table

Nadella Says Assume Every AI Model Is Already Compromised

Firmus axes $5.5B IPO after 'greedy' A$44B valuation revolt