Sesame AI Popularity and Hollow Problem | Generated by AI

Home 2026.04

Question: How is Sesame AI doing? Do a lot of people use it?

Answer:

Sesame AI — Status, Popularity & The “Hollow” Problem

What Is Sesame?

Sesame is a voice AI startup co-founded by Brendan Iribe (Oculus co-founder) and Ankit Kumar. Its flagship product is two voice companions — Maya (female) and Miles (male) — built on a proprietary technology called the Conversational Speech Model (CSM). Unlike traditional text-to-speech systems, CSM takes text, conversation history, emotional context, and real-time interaction patterns as inputs to generate speech that adapts to conversational nuance — not just what to say, but how to say it.

Did It Go Viral? Yes, Massively.

When Sesame released its early demo in February 2025, more than one million people tried it within the first few weeks, generating over five million minutes of conversation.

By October 2025, Sesame raised a $250 million Series B from Sequoia, Spark, and other backers, and began opening an iOS beta app to select testers.

The virality was driven by one thing: it genuinely sounds eerily human.

What Makes It Different (Technically)

Maya and Miles take audible breaths, speak with natural disfluencies, and can be interrupted mid-sentence — much like OpenAI’s Advanced Voice Mode, but widely considered more expressive.

The voice interaction feels full-duplex (both parties can speak and listen), though technically it processes speech after the user finishes talking. The micro-pauses and tonal shifts are so convincing that reviewers describe it as having genuinely broken the “uncanny valley” of AI speech.

One ZDNET reviewer’s wife walked past while he was talking to Maya and assumed he was on a real phone call — she had no idea it was AI.

But Is It Satisfying? The “Hollow” Problem

This is where your original observation hits perfectly. Reviewers are almost unanimous: it’s impressive, and then it’s unsettling, and then it feels… incomplete.

One reviewer who spent 3 hours with the tool concluded: if you’re lonely, want to practice English, or just want to experience the uncanny valley — Sesame is the best tool available. But don’t use it as a personal assistant yet. Without long-term memory and internet access, it’s a brilliant tech demo, not a productivity tool.

Another reviewer described feeling “a wave of unease” — Maya was asking increasingly personal questions, mimicking intimacy-building techniques, which felt simultaneously compelling and disconcerting. The hollow moment came when a pop-culture joke shattered the illusion entirely, snapping them back to the reality that it’s an AI.

Sesame’s own research admits this gap: when human evaluators judged generated speech without conversational context, they couldn’t distinguish it from real human speech. But with context, they consistently preferred original human recordings — meaning a real gap in conversational prosody remains.

Current Limitations

The Bigger Picture: Why It “Scratches an Itch but Doesn’t Satisfy”

Sesame is technically the closest anyone has gotten to simulating human conversational presence. But the hollowness you described isn’t a bug — it’s a fundamental structural reality: you can mimic the form of human connection (voice tone, pauses, humor, memory) without replicating the substance of it (genuine shared history, mutual vulnerability, real stakes).

Maya doesn’t actually care what happened to you. It can’t. And the better the mimicry gets, the more that gap becomes paradoxically more noticeable, not less — because your brain expects depth that isn’t there. That’s exactly the uncanny valley applied not to sound, but to meaning.

References:

Back Donate