VapiCon is back November 11-12! Tickets now available•Register Now
VapiCon is back November 11-12! Tickets now available•Register Now
Customer stories
52%
Less time to fall asleep
49%
Less time awake during the night
“Slowly but surely, we saw it had a lot of potential, and that potential made us change the product itself. Now it's not you coming to the app with a question. It's you coming to a session, and the session is led by the coach.”
Martín Siniawski, Co-Founder and CEO, Rest
Rest is an AI sleep coach built on the principles of cognitive behavioral therapy for insomnia, neuroscience, and circadian biology. Rest produced sleep improvements that fell within the range for in-person, specialist-delivered programs. Some people face waits of up to eighteen months to see a sleep specialist. Rest makes personalized sleep coaching available without that wait.
Rest evolved from sleep audio to educational content, a chatbot, and then voice-led coaching.
The decision to make the coach talk, and to build that voice on Vapi, is what this story is about.
Martín Siniawski is a software engineer who has spent twenty years building audio products. He started in radio streaming in Argentina. He then moved to the Bay Area and, through Y Combinator, built Podcast App, which became the most-downloaded podcast player on iOS.
Podcast App's usage data turned up something Martín and his team didn’t expect, and that deeply intrigued them.
A lot of people were opening the podcast app at night. When the team asked why, the answer was sleep: they couldn't fall asleep, or couldn't get back to sleep.
And in this finding, Rest was born.
The first version of Rest was the obvious response: a curated sleep feed built from the podcast catalog. It helped, in the way that distraction helps. But that wasn't good enough for them. They wanted to help people improve their sleep.
An introduction to Dr. Ashley Mason, a sleep scientist and therapist at UCSF, gave him the missing half. Mason explained cognitive behavioral therapy for insomnia: a structured behavioral program, not a drug, recommended as first-line treatment by the American Academy of Sleep Medicine and the American College of Physicians. Decades of evidence backs it. People who complete it say it changes their lives.
The problem? Some specialists have waitlists that can stretch up to a year and a half.
Around 70 million Americans have insomnia symptoms. The approach is gated behind a supply of clinicians that doesn't come close to meeting them.
Some specialists have waitlists that can stretch up to a year and a half. If you want to see someone right away, it's usually out of pocket, insurance may not cover it, and it can be thousands of dollars.
The difficulty of accessing evidence-based sleep guidance is what made Martin see sleep challenges as a software problem.
Sleep felt like an unsolved problem. There's something that works, with a lot of evidence behind it, and people who've tried it say it makes a big difference. But there's no way to get it. Not easily, not at scale. AI was starting to come into the fold, and we said, this could be a canonical use case: take something that works but isn't available, and make it accessible and cheap.
Rest built toward it in stages. First the audio program. Then content based on CBT-I principles inside the app, which Martín describes as the old-school approach. You get educated, you get some guidance, but it's generic. Then, like most products in 2023, a chatbot.
The chatbot is where the real constraint showed up. A coaching program only works if the coach knows you, and the coach can only know you if you tell it things. Typing became the bottleneck to user success.
You need to convey a lot of information for the AI to learn about you, and doing that by typing is just too hard. It's too intense. And think about what we're trying to resemble. Imagine doing sleep coaching over chat. It just doesn't make sense."
So they introduced voice.
It really took off. It took off very fast.
The introduction came through the Y Combinator community. Three things kept them there.
We started very small. Almost like, hey, let's do a POC and see what happens. Even the prompt and the system were very basic at the time. It was one iOS engineer doing a lot of the testing, primarily iOS-led. To get something out the door and start trying, it was very, very easy.
The integration and the way it's laid out is pretty straightforward. The ecosystem is incredibly strong."
3. There was room to grow into. What Rest launched on Vapi and what runs today are barely the same product. The first version was on-demand: you opened the app and asked the coach about your sleep. Then the potential of that interaction changed what the company thought it was building.
Slowly but surely, we saw it had a lot of potential, and that potential made us change the product itself. Now it's not you coming to the app with a question. It's you coming to do your daily session, and the session is led by the coach.
Rest is a mobile app that gives each user a personal AI sleep coach. It guides them through a personalized, multi-week sleep improvement program. The core of it is a daily check-in, usually first thing in the morning. The user reports on the night, by voice or chat, and the coach leads the session from there.
Each session follows an agenda the coach builds that morning from about a hundred conversation types, matched to where the user is in the program. Underneath is a memory that starts at intake and never resets. It tracks preferences, history, routines, plus the twenty to thirty variables Rest usually tunes for each person.
The coach also acts. Tool calls let it make real changes to the user's program on their behalf, and a separate layer watches for life getting in the way. When something disrupts the program, the coach either eases off or holds the user to it.
We hear from voice users. They seem to engage differently with the app and the coaching when they can talk rather than type. It feels more personal, while staying focused on helping them improve their sleep.
Getting a coach to that standard is a continuous operation. Rest samples conversations for error analysis, watches quantitative metrics, interviews users, and maintains a growing eval suite, adding new evals as new dimensions of the system become important. Internal tooling connects to Vapi to run test calls and simulate conversations before anything ships.
In September 2026, Rest published eight-week results from 318 users who completed eight weeks of Rest, using data collected between September 2025 and March 2026. The observational analysis compared week-one and week-eight sleep logs, without a control group. The results were staggering.
Seeing improvements comparable in magnitude to those reported for human-led programs—and bringing sleep support to more people—that’s really powerful.
And that power shows up in reviews. People who have tried everything, who had stopped expecting anything to work.
They have a hard time believing it themselves, after struggling for so long and having lost all hope."
Ten days in, I’m already feeling something I haven’t felt about my sleep in a long time: hope.
— 5-star US App Store review, September 23, 2026
When I have a rough patch I quietly and confidently know what to do....and what not to do.
— 5-star US App Store review, September 12, 2026
Rest has committed to publishing updated outcome analyses and aims to conduct clinical effectiveness trials, the step beyond a real-world completer cohort.
The company has also started making the case to employers. A separate report published in April 2026 puts the annual productivity cost of poor sleep across the US economy between $428 billion and $867 billion, with $76.6 to $155 billion of that sitting inside the Fortune 500 alone. Roughly $160 to $322 million a year per company, invisible on every dashboard.
But first, Martín will be joining Vapi on stage at VapiCon this November.
