{
"$type": "site.standard.document",
"content": {
"$type": "blog.pckt.content",
"items": [
{
"$type": "blog.pckt.block.text",
"plaintext": "There's a particular moment in every AI voice interaction that breaks the spell. The assistant finishes its perfectly accurate sentence, and instead of responding naturally, it waits — silently, patiently, like a customer service machine — for you to give it the next command. The rhythm is wrong. It feels like filling out a form, not holding a conversation."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "Sesame AI, the startup co-founded by Oculus co-founder Brendan Iribe, has spent the last several years obsessing over that exact problem. This week, after more than a year of development and a beta that reached over a million users, the company launched its iOS app publicly across 39 countries. Its pitch is deceptively simple: AI that actually sounds like it's listening."
},
{
"$type": "blog.pckt.block.text",
"facets": [
{
"features": [
{
"$type": "blog.pckt.richtext.facet#bold"
}
],
"index": {
"byteEnd": 373,
"byteStart": 352
}
}
],
"plaintext": "\nThe Sesame founding team knows something about transformative interfaces. Iribe and several co-founders helped build Oculus, the VR headset that Meta acquired for $2 billion in 2014. That experience — chasing immersive presence, bridging the gap between digital and physical experience — clearly informs what Sesame is building today. Backed by a $250 million Series B, the company's long-term ambition is lightweight AI-enabled eyewear, expected in 2027. The app, is a proving ground for that vision — getting the voice right before building the hardware around it."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "This tension, articulated in Sesame's own launch materials, captures the central design problem the company is trying to solve. Their answer is a system that thinks and talks simultaneously — running fast retrieval and search in parallel while speaking, weaving up-to-date information into responses mid-sentence rather than pausing to look things up."
},
{
"$type": "blog.pckt.block.text",
"facets": [
{
"features": [
{
"$type": "blog.pckt.richtext.facet#bold"
}
],
"index": {
"byteEnd": 73,
"byteStart": 40
}
}
],
"plaintext": "The heart of Sesame's technology is its Conversational Speech Model (CSM), first unveiled in February 2025. Unlike conventional text-to-speech pipelines that convert words into audio as a finishing step, CSM was designed from the ground up as a unified, end-to-end system. The architecture is technically distinctive in several ways."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "This joint processing approach — rather than a sequential pipeline — is what enables the model's most striking quality: real-time contextual adaptation. The system can modulate tone and pacing mid-sentence in response to conversational cues it's still receiving. It doesn't decide how to sound before it speaks; it figures it out as it goes, just like people do."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "The app launches with four named agents — Maya, Miles, Simone, and Charlie — each with distinct personalities and conversational styles. This isn't cosmetic branding. The multiple-agent approach lets Sesame explore how different voice personas build different kinds of rapport, which matters enormously for the company's longer-term vision of a companion that's genuinely woven into daily life."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "The app supports live search, note-taking, conversation summaries, and an incognito mode — all accessible through voice. The full experience is currently free, a deliberate choice to prioritize adoption and habit formation over early monetization. The real product, after all, is the relationship between user and agent, and that takes time to develop."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "The conversational AI space is extraordinarily crowded. ChatGPT, Gemini, Claude, and a dozen others all offer voice modes with improving fidelity. The fundamental difference is philosophical before it's technical. ChatGPT and Gemini are text-native systems that learned to speak. Sesame is voice-native from the architecture up. That distinction compounds: every design decision, from latency optimization to personality consistency, flows from a different set of first principles."
},
{
"$type": "blog.pckt.block.text",
"facets": [
{
"features": [
{
"$type": "blog.pckt.richtext.facet#italic"
}
],
"index": {
"byteEnd": 132,
"byteStart": 129
}
}
],
"plaintext": "Reviewers who tested early versions have described the experience in unusually vivid terms, drawing comparisons to Spike Jonze's Her— the 2013 film in which a man falls in love with an AI operating system. That's either a powerful endorsement of the technology's realism or a useful caution about where hyper-naturalistic AI companions can lead, depending on your perspective."
},
{
"$type": "blog.pckt.block.text",
"facets": [
{
"features": [
{
"$type": "blog.pckt.richtext.facet#bold"
}
],
"index": {
"byteEnd": 108,
"byteStart": 90
}
}
],
"plaintext": "The iOS app is very explicitly a stepping stone. Sesame's published roadmap points toward agentic capability — AI that doesn't just think with you, but takes actions on your behalf — followed by the 2027 eyewear launch that would embed these voice agents into hardware you wear throughout the day. The goal is ambient AI: not something you open on your phone, but something that's simply present."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "This is a high-stakes bet. The history of AI wearables is littered with expensive failures. But Sesame's approach — proving the voice experience first, building the habit, then adding hardware — is meaningfully more grounded than the \"build the device and hope\" strategy that has tripped up previous attempts."
},
{
"$type": "blog.pckt.block.text",
"plaintext": "For now, the most honest answer is that Sesame has built something that many people describe as genuinely different to use — an AI that sounds less like software and more like someone on the other end of a phone call. Whether that quality can anchor a daily habit, a hardware business, and a new category of personal computing remains to be seen. But it's a compelling place to start."
},
{
"$type": "blog.pckt.block.text",
"plaintext": ""
}
]
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreiehzhj5wqhoyr4boarsuon5n42htqa2sz53x35jsnpffi7v4il4mu"
},
"mimeType": "image/png",
"size": 53882
},
"description": "There's a particular moment in every AI voice interaction that breaks the spell. The assistant finishes its perfectly accurate sentence, and instead of responding naturally, it waits — silently, patiently, like a customer service machine — for you to give it the next command. The rhythm is wrong. It feels like filling out a form, not holding a conversation. Sesame AI, the startup co-founded by Oculus co-founder Brendan Iribe, has spent the last several years obsessing over that exact problem. Th...",
"path": "/the-app-that-wants-to-sound-human-gezx55y",
"publishedAt": "2026-05-29T17:58:22+00:00",
"site": "at://did:plc:5veoox46kt7lc3cgkx7ro6l2/site.standard.publication/3mmwtqiwv3e4e",
"tags": [
"Tech"
],
"textContent": "There's a particular moment in every AI voice interaction that breaks the spell. The assistant finishes its perfectly accurate sentence, and instead of responding naturally, it waits — silently, patiently, like a customer service machine — for you to give it the next command. The rhythm is wrong. It feels like filling out a form, not holding a conversation.\nSesame AI, the startup co-founded by Oculus co-founder Brendan Iribe, has spent the last several years obsessing over that exact problem. This week, after more than a year of development and a beta that reached over a million users, the company launched its iOS app publicly across 39 countries. Its pitch is deceptively simple: AI that actually sounds like it's listening.\nThe Sesame founding team knows something about transformative interfaces. Iribe and several co-founders helped build Oculus, the VR headset that Meta acquired for $2 billion in 2014. That experience — chasing immersive presence, bridging the gap between digital and physical experience — clearly informs what Sesame is building today. Backed by a $250 million Series B, the company's long-term ambition is lightweight AI-enabled eyewear, expected in 2027. The app, is a proving ground for that vision — getting the voice right before building the hardware around it.\nThis tension, articulated in Sesame's own launch materials, captures the central design problem the company is trying to solve. Their answer is a system that thinks and talks simultaneously — running fast retrieval and search in parallel while speaking, weaving up-to-date information into responses mid-sentence rather than pausing to look things up.\nThe heart of Sesame's technology is its Conversational Speech Model (CSM), first unveiled in February 2025. Unlike conventional text-to-speech pipelines that convert words into audio as a finishing step, CSM was designed from the ground up as a unified, end-to-end system. The architecture is technically distinctive in several ways.\nThis joint processing approach — rather than a sequential pipeline — is what enables the model's most striking quality: real-time contextual adaptation. The system can modulate tone and pacing mid-sentence in response to conversational cues it's still receiving. It doesn't decide how to sound before it speaks; it figures it out as it goes, just like people do.\nThe app launches with four named agents — Maya, Miles, Simone, and Charlie — each with distinct personalities and conversational styles. This isn't cosmetic branding. The multiple-agent approach lets Sesame explore how different voice personas build different kinds of rapport, which matters enormously for the company's longer-term vision of a companion that's genuinely woven into daily life.\nThe app supports live search, note-taking, conversation summaries, and an incognito mode — all accessible through voice. The full experience is currently free, a deliberate choice to prioritize adoption and habit formation over early monetization. The real product, after all, is the relationship between user and agent, and that takes time to develop.\nThe conversational AI space is extraordinarily crowded. ChatGPT, Gemini, Claude, and a dozen others all offer voice modes with improving fidelity. The fundamental difference is philosophical before it's technical. ChatGPT and Gemini are text-native systems that learned to speak. Sesame is voice-native from the architecture up. That distinction compounds: every design decision, from latency optimization to personality consistency, flows from a different set of first principles.\nReviewers who tested early versions have described the experience in unusually vivid terms, drawing comparisons to Spike Jonze's Her— the 2013 film in which a man falls in love with an AI operating system. That's either a powerful endorsement of the technology's realism or a useful caution about where hyper-naturalistic AI companions can lead, depending on your perspective.\nThe iOS app is very explicitly a stepping stone. Sesame's published roadmap points toward agentic capability — AI that doesn't just think with you, but takes actions on your behalf — followed by the 2027 eyewear launch that would embed these voice agents into hardware you wear throughout the day. The goal is ambient AI: not something you open on your phone, but something that's simply present.\nThis is a high-stakes bet. The history of AI wearables is littered with expensive failures. But Sesame's approach — proving the voice experience first, building the habit, then adding hardware — is meaningfully more grounded than the \"build the device and hope\" strategy that has tripped up previous attempts.\nFor now, the most honest answer is that Sesame has built something that many people describe as genuinely different to use — an AI that sounds less like software and more like someone on the other end of a phone call. Whether that quality can anchor a daily habit, a hardware business, and a new category of personal computing remains to be seen. But it's a compelling place to start.",
"title": "The app that wants to sound human",
"updatedAt": "2026-05-29T18:22:38+00:00"
}