The Oculus co-founder released an iOS app this week that wants you to forget you're talking to a bot. Not through perfect mimicry—Sesame's four AI agents don't pretend to be human. Instead, they try something perhaps more radical: letting you ramble, change your mind mid-thought, and interrupt yourself the way people actually do when they're thinking out loud.
Sesame went public on May 28, 2026, launching in 39 countries as a free preview. The app gates access with a waitlist (how long depends on server capacity and demand), but once you're in, you get four distinct voice agents—Maya, Miles, Simone, and Charlie—each with its own personality and memory. They search the web in real-time while you talk. They can pivot if new information arrives mid-sentence.
It's conversational AI designed for the way conversations actually happen, not the stilted back-and-forth of typical voice assistants. Whether that's what people want—or will pay for—is the experiment Iribe's now running in public.
The Mechanics: Parallel Search, Live Pivots
Most voice assistants follow a predictable pattern. You ask a question. They think. They answer. Repeat.
Sesame's agents work differently, running multiple parallel searches during a single exchange and weaving results into responses on the fly. If something relevant streams in while Maya (one of the four agents) is already speaking, she'll course-correct mid-answer. TechCrunch described this approach as "rethinking chatbots toward fluid conversation even when the AI needs time to think."
The technical foundation comes from work Sesame detailed in a February 27, 2025, technical post on what it called a Conversational Speech Model—a system trained on roughly one million hours of public audio, built for low-latency, contextual responses. The Verge, which saw an early demo of the Maya agent around that time, called it "genuinely fun" and more natural than the competition.
That's the pitch, anyway. The reality involves 30-minute call limits for logged-in users (five minutes if you're not signed in), conversation summaries sent after each session, and support for English only at launch. An Android version is coming, the company said in a blog post the day before launch, though no specific timeline was offered.
Four Voices, One Curiosity Engine
Each agent—Maya, Miles, Simone, Charlie—maintains separate memory across voice and text conversations, building context over time. Users can switch between them depending on the task or their mood. Want a more analytical take? Try Miles. Prefer something lighter? Maybe Charlie.
The company frames the design as a "curiosity engine," balancing quick conversational turns with deeper thinking when warranted. The interface includes search cards with images, notes, reminders, and something called "Deep Dives" for more thorough exploration of a topic. You can text with the agents too, though voice is clearly the primary experience.
An Incognito Mode lets users access prior context without saving new interactions—for those moments when you'd rather the conversation stay ephemeral.
The Oculus Alumni Network
Iribe's pedigree here matters. He co-founded Oculus, the VR company Facebook acquired in 2014 for $2 billion, and his new venture pulls from that same talent pool. Nate Mitchell, another Oculus co-founder, runs product. Hans Hartmann, former Oculus COO, holds the same title at Sesame. Iribe launched the company in 2023 alongside Ankit Kumar (ex-CTO at Ubiquity6) and Ryan Brown (an Oculus engineering manager who later worked at Reality Labs).
The startup raised a $250 million Series B announced on October 21, 2025, led by Sequoia Capital with Spark Capital participating. Andreessen Horowitz, Spark, and Matrix Partners had backed earlier rounds. Sequoia's investment note at the time cited usage from the February 2025 demo: more than one million people, over five million minutes of conversation in the first few weeks. Those figures are over a year old now and don't reflect current scale, but they signaled early interest before the public launch.
Sesame operates from San Francisco, with additional offices in Seattle and New York. Angela Gayles, a longtime Facebook and Meta executive, joined to oversee operations.
Free Now, Eyewear Later
The iOS app—which requires iOS 18.0 or higher—is free for now. No ads, no data selling, the company promises. What it hasn't announced: subscription pricing or when that might kick in.
The bigger ambition extends beyond software. Sesame is targeting intelligent eyewear for 2027, a timeline reiterated in launch materials. That's where Iribe's hardware background becomes relevant—the long game isn't just conversational AI on your phone, but agents you can interact with throughout your day via glasses or some other wearable form factor.
TechCrunch noted the iOS launch feels like groundwork for a future where these agents don't just answer questions but take actions: booking travel, managing calendars, that sort of thing. The agentic behavior isn't there yet, but the conversational foundation—such as it is—now exists.
Early Signals, Long Bets
The App Store listing shows Sesame with a 4.9 average rating from 303 reviews as of early June. Those numbers shift quickly and matter less than sustained usage patterns, which remain opaque. What's clearer: the team is iterating fast. Version updates rolled out frequently in May—22.0 on the 14th, 23.0 on the 20th, then 24.0 and 24.1 in late May. That's the cadence of a company refining in public, adjusting based on what early users encounter.
Whether voice-first AI agents become a genuine interface shift or remain a curiosity for early adopters depends on execution, yes, but also on whether people actually want to talk to their devices this way. Plenty of sophisticated voice tech has launched with fanfare only to collect dust—remember Siri's early promise?
Sesame's pedigree buys it credibility and runway. The free preview offers a low-stakes test before the hardware arrives. For now, it's available to try, waitlist permitting.
Which raises the real question: not whether the technology works, but whether conversational AI solves a problem people know they have.
