BUTTERFLY NETWORK’S NEXT CHIP: JOE DEVIVO RETURNS ON AI, ULTRASOUND & THE PATH FROM PROTOTYPE TO PRODUCT

The Harry Glorikian Show

Butterfly Network – Joe DeVivo, CEO

For July 29, 2026

Three years after his first appearance, Butterfly Network CEO Joe DeVivo returns to catch Harry up on the roadmap. Butterfly’s ultrasound-on-a-chip has evolved from a single handheld device into a foundational semiconductor platform that third parties can build on — most dramatically, Midjourney’s full-body scanner, which rings 40 Butterfly chips around the body for whole-body tomographic imaging. Joe and Harry dig into what’s real today versus what’s coming, the physics of ultrasound, the path from demo to FDA clearance, and the business logic of “multiple shots on goal.”

In This Episode

The chip as a platform: opening Butterfly’s technology to third-party developers (the “Nvidia moment”)
Midjourney’s full-body scanner: tomography, 400 teraflops of data, and the road to a one-minute scan with the Apollo chip
Real-world impact: 1,500 midwives trained in Kenya and South Africa, plus a new FDA-cleared gestational-age AI tool
Partners beyond Midjourney: vascular robotics, fatty-liver diagnosis, and brain-computer interfaces (including “silent speech”)
Physics Q&A: why 360-degree imaging solves ultrasound’s problems with air and bone
The hospital shift: capturing “ghost” scans, Compass AI, and at-risk skilled-nursing deployments
The 10-year vision: imaging that comes to the patient instead of the patient going to imaging

Chapters

00:01 A roadmap from three years ago
00:31 Midjourney unveils a full-body scanner
01:31 Joe DeVivo returns
02:59 What shipped: the chip becomes foundational
04:52 Opening the platform to third parties
06:34 Kenya & South Africa: midwives and the FDA-cleared gestational-age app
11:11 FDA rigor and validating AI
14:52 Butterfly Garden, Butterfly Embedded, and the nine partners
16:53 Brain-computer interfaces and silent speech
22:00 The Midjourney deal: origin and reaction
24:03 Tomography explained
28:12 What 40 chips can see and the data challenge
29:33 The roadmap to a one-minute scan; the Apollo chip
32:43 AI looking for a physical carrier
34:31 Prototype to product: milestones to clearance
38:28 Wellness first, then regulatory clearances
39:03 The physics of air and bone
41:45 The investment thesis: multiple shots on goal
46:26 The business today: Compass AI and health-system deployments
49:54 Heart-failure monitoring and at-risk skilled nursing
53:49 The 10-year view: imaging comes to the patient
57:18 Why early detection matters
59:50 Closing

About the Guest

Joe DeVivo is CEO of Butterfly Network and has spent 35 years in medical devices.

Full Transcript

Harry Glorikian (00:01) Three years ago, I had Joe De Vivo on this show. He runs the Butterfly Network, the company that put an entire ultrasound machine onto a semiconductor chip. And he told me something in that conversation that I’ve quoted in my talks ever since. He said Butterfly’s first act was like strapping a combustion engine onto the back of a horse. The engine was the chip, the horse was the handheld probe. The vehicle, he said, was still coming.

Harry Glorikian (00:31) Well, on June seventeenth, mid journey, and yes, I do mean the image company your kids use to make dragon pictures, walked on stage in San Francisco and unveiled a full body scanner. You step onto a platform, you get lowered into a pool of water, and a ring of sensors images your entire body on the way down. Inside that ring, 40 of butterflies chips, butterfly stock jumped

Harry Glorikian (01:00) More than 50% in that day. Now, this announcement generated a lot of excitement and a lot of questions. And Joe has spent 35 years in medical devices, which makes him exactly the right person to help us separate the two. So he’s back, and we’re going to talk about what’s real today, what’s coming, and what happens when images or imaging stops being a place you get sent to and starts being something that’s just there.

The Harry Glorikian Show (01:31) Joe, welcome back to the show.

Joseph DeVivo (01:34) Good to see you, Harry.

The Harry Glorikian Show (01:36) I you know, I I was just telling you a minute ago, like I I was like, I went back and I was like, let me look, let me review, let me review what what we talked about the last time we talked. And in my mind, I’m like, it wasn’t that long ago. And I’m thinking, it was 2023. I was like, my God.

Joseph DeVivo (01:53) Yeah. It’s an eternity.

The Harry Glorikian Show (01:57) Like, I mean, I think I’ve been so stuck in the AI world that that it’s just like time is going by and I’m not realizing that the time is like flying by at this point. But

The Harry Glorikian Show (02:06) Look, I I gotta encourage people that if you if you haven’t had a chance to listen to those conversations from 23, it’s totally worth going back into the archive and and sort of listening to them. But I went back and sort of thought about it and I thought, he laid out a roadmap for me. And and so let me re repeat this back to you, and you tell me where we are at this point. Is you told me the next probe, now going back to 23, would have an auto scan feature.

The Harry Glorikian Show (02:35) Take slices like an MRI so a less experienced person could place it and press a button. You told me about a home monitoring version, a device the size of your AirPod case for heart failure patients. So catch me up, walk me down the roadmap, what shipped, what’s coming, what surprised you along the way?

Joseph DeVivo (02:59) Well think I think the biggest thing since the last we spoke that that’s kind of come to fruition that includes all the things that you just mentioned was we learned that the semiconductor that we use to create images for our device. We’ve learned that that semiconductor is foundational. you know we’ve learned that

Joseph DeVivo (03:25) We solved the problem for ourselves. The the vision of our founder was to create a device was to democratize medical imaging. Two-thirds of the world doesn’t have a radiology department to just wal to go into and have a scan. You know, rural sites, third world countries, they don’t have access. And and we all know that early detection of any ailment, you’re gonna be able to have an increased probability of success the earlier you find something. And so

Joseph DeVivo (03:54) You know, we’ve created this semiconductor to allow one handheld ultrasound device to image the entire body and to have the different arrays, different algorithms and different frequencies necessary for the heart, the lungs, you know, for the eyes, for the carotids, the peripheral vascular, MSK, et cetera. So th but we it it costs us three hundred million dollars to make our first chip. And and we’ve been on this journey of

Joseph DeVivo (04:23) making better processors so our image quality gets better, very similar to to digital photography. Like each new digital camera, you know, the the more compression it had and resolution the more you know the more megapixels they were able to fit on a screen. you know, finally people started replacing film cameras with digital cameras. Well, we’ve learned that that chip that we created can be used by other companies.

Joseph DeVivo (04:52) We learned that, you know, that what we were doing was so we solved the problem for ourselves that other people had. So we opened our platform two years ago, Harry, and we allowed third party companies to start building products using our chip. And you know, from from from nineteen ninety-three to two thousand and five, Nvidia was a video game.

Joseph DeVivo (05:19) There was nothing in their business plan that said they were getting into AI, that they were getting into data centered compute. But in 2006 they made a profound decision, and that was to build a software platform that would allow third-party developers to be able to change the settings and maximize the compute for different use cases. And so we’re replicating that model. We have launched a developer software that is now allowing for third parties to do everything that we said there that they

Joseph DeVivo (05:47) You know, we are using it, but we’re now allowing others to be able to do a fan scan without moving the device or a slice to be able to capture multiple images. And probably the you know, the most exciting manifestation of this is Mid Journey, who has taken forty of our ultrasound probes, put them in a circle, and are able to image the entire body using

Joseph DeVivo (06:16) you know, these capabilities that we just talked about. And so we you know, we are growing as a point of care ultrasound company, but we are now growing a business as a semiconductor based ultrasound company that’s empowering some really, really cool stuff.

The Harry Glorikian Show (06:34) So I st I still want to go back though and see like you know we’re we’re with the premise of like really bringing this to the masses. I think you told me back in 23, we were talking about Kenya, and you you told me people, you know, like I think it was the Gates Foundation, bought a thousand probes, 500 midwives trained in two weeks.

The Harry Glorikian Show (07:03) And I think you gave me a stat on the air at that time that 377 of the 500 were actively scanning, and 80% said they caught some sort of high-risk issue that got resolved. Now that was almost three years ago. Is that still happening? Are you guys still pushing that envelope? You know, what what can you tell us around that?

Joseph DeVivo (07:26) Well very much so. And so we now, as you said, we we put a thousand five hundred we trained five hundred midwives in Kenya. we then trained through our partners Gussi, five hundred midwives in South Africa. So then there’s a thousand probe, thousand midwives. And actually that population, we because all of our devices aren’t networked in the cloud, we have access to

Joseph DeVivo (07:52) The scan data, but also, you know, the performance metrics. So we know how many scans are being done a day, what the presets are. And that group is probably some of the highest volume scanners we have in the world. Because every day they’re doing scans to to help high-risk pregnancies and in all pregnancies. And it’s and that is continu we haven’t seen a slowdown of scans. We’ve actually seen an increase in scans among that population.

Joseph DeVivo (08:21) And then recently we received an FDA approval for a new AI tool that does gestational age. So you I don’t know if you’ve been in an obese suite, you know, with someone having a a baby, but but basically when w what the ultrasonographer is doing is measuring the length of the femur, measuring the length of the tibia, measuring the diameter of the head.

Joseph DeVivo (08:52) And then based upon those measurements, they come up with what the age of the child is. But in rural communities and in third world countries, a lot of times women don’t know actually when they conceived. And that is instrumental in their health because the for the as the the baby ages, both the mother and the child have different healthcare requirements.

Joseph DeVivo (09:20) At different phases of their gestation. And not knowing the actual age is detrimental to the process. So we came out with an app called the Just a Generic Gestinational Age app. And instead of having to go and manually measuring each of these body parts and then calculating them, it automatically sweeps through the body.

Joseph DeVivo (09:45) And then as it’s sweeping through the body, it’s cat it’s identifying the anatomy, simultaneously measuring the anatomy, and then within two minutes generates the age for the baby. So what’s so incredible about our architecture is just like an iPhone, where, and this is not typical in medical. When I was running other healthcare companies, if I wanted to change the software in my device, I’d have to send the rep in the field with a USB stick. You know what I mean? I mean, it’s not networked.

Joseph DeVivo (10:16) But now we can push an update to all thousand of those nurses. Give them a new tool, a gestational age app. And then they can use this not only to determine fetal position, but then they can help women understand what their gestational age is. So that gets them the best care possible. And you can do it by pushing a software update to all the thousand devices overnight. So so the delivery of that.

Joseph DeVivo (10:42) is easy. So that w we’re going through a very significant evolution. The ability of delivering these sweeping tools, like we mentioned in AI tools, the and and it’s a flywheel. Once you have a user understand a an initial use case, then their comfort level increases and adding the next use case, next AI tool, next scan really be is it it it’s a it’s a a a reduced effort which with each new innovation.

The Harry Glorikian Show (11:11) And it seems like,

Joseph DeVivo (11:12) Mm.

The Harry Glorikian Show (11:12) you know, the FDA is starting to catch up or open the opportunity for.

The Harry Glorikian Show (11:21) more applications being cleared, where, you know, instead of having to necessarily go after them one by one, you might be able to go after a a number of them under one application from a few companies that I’ve seen so far. And I think

Joseph DeVivo (11:36) Well

The Harry Glorikian Show (11:37) that’s a big shift.

Joseph DeVivo (11:38) yeah, you know, I I think you know I’ve been been doing this a long time. I’ve I’ve worked for a lot of different companies in different markets and have been on on the good side and the bad side of FDA. they’ve been on I’ve had my name on a consent decree with Department of Justice and and I I’ve you know I I’ve gotten a lot of great clearances also. I have a profound respect for FDA. I think they’re

Joseph DeVivo (12:05) I wish I you know, not not to pull out an RFK thing, but I I wish I I I wish our food industry had the type of rigor, you know. I I know the F in the FDA is not rigorous enough. I you know, if we’re gonna change our food supply, they should go through the same type of effort that we change our medical devices. You know, I would always joke that, you know, as as a medical device company, h here we are, you know, selling sophisticated devices.

Joseph DeVivo (12:35) To sophisticated users. And we need to be told exactly what claims we can have and what we can say to that sophisticated population. But Tony the Tiger can tell my kids that, you know, this sugar cereal is great without any consequence. But to FDA, though, I have tremendous respect for FDA. I actually think that.

Joseph DeVivo (12:59) You know, whether it’s a piece of plastic, whether it’s a wire, whether it’s a software app or whether it’s an AI tool, I think companies have to, you know, be able to prove that whatever they’re doing, whatever information they’re giving to a patient, whatever they’re doing, and and whatever they’re educating a user, that they need to validate that whatever they’re saying is true. and and and

The Harry Glorikian Show (13:22) a hundred percent. A hundred percent.

Joseph DeVivo (13:24) I think sometimes d we we we

Joseph DeVivo (13:27) like companies complain about the difficulty of getting AI approved because they don’t think they need to do the type of validation that you would a medical device. But if if your AI is telling someone to alter their care or telling someone, telling a doctor to do something differently, it better be right. And that’s where I that’s where I think the FDA gets it right is, you know, they’re they’re not allowing for there to be shortcuts just because the technology changes.

Joseph DeVivo (13:54) It’s all about patient safety and what the consequence is. And and I do think the industry is coming into a a better understanding of how to do this. And the FDA is certainly streamlining and and understanding also. But we know that gestational app that we were where that we received approval from from the FDA did have a very large clinical data set. It was dealt with as almost, you know, a device or like a device showing that.

Joseph DeVivo (14:23) when we received these types of measurements and we did this type of work, that it was accurate and that it was verified. And it’s just not just because it’s software that it it doesn’t have consequence or it’s AI. Actually you can even argue it’s a higher consequence. So you know, I I do think that the those who produce products and those who approve them, I think they understand things better. But I think people in the beginning were complaining because they just weren’t doing

The Harry Glorikian Show (14:52) So, Joe, you talked about Midjourney a little earlier, but but you know, that I think that was based on the, you know, we when we talked back then, I think you were just launching what I think you called butterfly garden and butterfly embedded. And I’m assuming that that that Midjourney is part of that same strategic direction you guys took. But I think I saw that there’s like nine partners as of April.

The Harry Glorikian Show (15:20) So besides mid journey, are there other ones that you can talk about that you

Joseph DeVivo (15:25) Yes.

The Harry Glorikian Show (15:26) think are exciting?

Joseph DeVivo (15:28) Yes, so so you know, we we have the only semiconductor in the in the world that can deliver ultrasound waves from a semiconductor. And what makes our semiconductor so unique is we have th 9,000 sensors in a rectangle on top of of mem sensors and it sits on top of a CMOS wafer. And that wafer allows us to control each one of those sensors individually. Like on this screen you have pixels, and each pixel the computer is

Joseph DeVivo (15:58) you know, programming at a different color to create an image, we can do that to create a very specific type of ultrasound wave. And and what that does is developers love it because they’re able to then with software customize signals and create signals that would be different than my than what you might use for a normal medical ultrasound. So we have, for example, Mendera, who is a a vascular robotics company.

Joseph DeVivo (16:25) And they use a bunch of features of our probe and our semiconductor to be able to pinpoint where they would like to put a needle in the body with tremendous precision and imaging. So they’re a partner, they actually have been a partner for a couple of years now. And they’re we’re very excited because in the next couple months, they’ll be going commercial with our platform. So that’s that’s a big one for us. we’ve announced a company called Sonic Insights that does fatty liver disease.

Joseph DeVivo (16:53) Diagnosis using ultrasound and it’s a precursor for some pretty ne necessary non-alcoholic fatty liver drugs that can help that can help people. one really fascinating area is brain computer interfaces. so Forrest Neurotech out of Caltech started to develop some some implantable devices that you can put in the brain.

Joseph DeVivo (17:19) That would then be able to determine neural activity and vascular activity. and we’re now, we probably have I think three of our nine partnerships today are brain computer interface companies. And they’re companies who are trying to in different ways stimulate the brain, learn and scan the brain. Because again, remember we talked about how we can fan that beam, right? That we can move that beam.

Joseph DeVivo (17:47) Well, if I put a chip in the brain and it can only look at one one dimension, then it’s not very impactful. But if I can use that chip and it can now scan the entire lobe of the brain or the entire brain itself. Like Harry, you should go to a website, Aleftneuro.com, A-L-E-P-H Neuro dot com. they’re one of our partners that just showed that they can do an an ultrasound of the brain through the skull.

Joseph DeVivo (18:16) using contrast media and completely map the vascular branches of the brain live. It’s absolutely insane and you should look at it. A left neural also not only do they create that image which will be used at some point in the future for a functional MRI type of image, but they’re also doing something called silent speech. So they take the probe and they put our ultrasound probe right here. And what they’re doing is that they’re imaging the tongue.

Joseph DeVivo (18:46) And so while they’re imaging the tongue, imagine if I start saying words, but there’s no sound. So say someone cannot speak sound, but they can mouth the words. Well, they’re creating these AI tools that can map the motion of the tongue to a word. And so someone can literally start.

Joseph DeVivo (19:09) And then the words come up on the screen and that that speech is then detected. So, you know, we’re we’re finding brain computer interface companies, vascular robotics companies, you know, scanning fatty liver companies. And we have you know, we have our earnings call coming up, so I can’t get too far ahead, but we’ll

The Harry Glorikian Show (19:26) Yeah.

Joseph DeVivo (19:27) we’ll we’re gonna be talking about some more.

The Harry Glorikian Show (19:31) I looked at the date and I was like, God, I can’t ask a few of these questions I wanna ask. But is the one where you press it against your throat, is that available already?

Joseph DeVivo (19:41) No, it’s i it it’s if you go into a leftneuro dot com you would see that it’s all in research right now. These are all companies that are doing this primary work. But what they’re showing, and they’re showing the w it’s like as each new company comes out and shows what they can do with ultrasound, because now they’re just programming. They’re pro they’re you know, they’re programming and they’re getting, you know, different arrays with different frequencies and megahertz and they’re they’re doing all these things and and and and

Joseph DeVivo (20:11) it’s opening up the possibilities of a whole new set of applications that that are really, really exciting. And and what it’s doing for Butterfly is, you know, we’re we we used to be a point of cure ultrasound company with a semiconductor. you know, over the next couple of years, we’re gonna be more of a an ultrasound semiconductor company with a point of cure ultrasound business. that other part of the business is really taking off.

Joseph DeVivo (20:38) are, you know, and we have 600 patents. TSMC, you know, makes our chips for us. So we have a very big moat and we make it really easy if someone wishes to do investigation in ultrasound with a semiconductor, they just buy our software, they buy some chips and they go. They don’t have to, you know, try to get through our IP or or whatever. but the moment they want to productize it, then they can we get into licensing and we get into purchase agreements and all kinds of stuff.

The Harry Glorikian Show (21:06) Yeah, I was thinking about the one on that you use on the throat. My one of my partners has had throat cancer a couple of times. And so I mean, he can he can speak, but you know, it’s sometimes it’s for me, I’m not my ears are not tuned to necessarily I have a I have a difficult time like hearing him. So I wonder I should I should turn him on to this in case he’s interested. But

Joseph DeVivo (21:28) Well

Joseph DeVivo (21:28) he he should definitely follow the field, because it’s something like when you get a breakthrough like this and it’s also it’s not like you’re taking a drug, right? It’s not like you’re taking something that has some consequence. You’re just putting an ultrasound probe here. It seems to me like that that silent speech idea probably would have a a pretty quick path to market if someone, you know, picked it up.

The Harry Glorikian Show (21:55) I just actually thought of something. We’ll talk about it when we’re done.

Joseph DeVivo (22:00) You wanna start a company, Ari? You could be our you could be our tenth partner.

The Harry Glorikian Show (22:02) You know,

The Harry Glorikian Show (22:06) you know, we’re all we’re always funding companies, we’re always trying to find new things, we’re always like trying to be on the bleeding edge, Joe. I was, you know, first thing I thought of when you were talking about the chip is, you know, are they using AI to sort of design the next chip, right? Just like, you know, OpenAI did with their jalapeno chip, like idea to

The Harry Glorikian Show (22:28) you know, taping it out in nine months, right? And and there’s no way a human did all that. That that there was an AI involved in helping modify that to the next level. So always trying to be on that bleeding edge of what’s going on. But let’s jump into like I want to go back November 17th of last year. You file with the SEC co-development and licensing agreement with Mid Journey, right?

The Harry Glorikian Show (22:59) 50 million up front, 10 million a year, milestones on top of that, like five year time, call it 74 million all in, plus they buy all the chips from you. And funny enough, like the market sort of shrugged until June 17th, when David Holtz gets on stage and shows the world what they’ve been building, and all of a sudden your j your stock just jumps, right? So

The Harry Glorikian Show (23:28) Take me back to the beginning. Like when Mid Journey for because I’m telling you, we’ve been using Midjourney, but not for anything like this, right? It’s to make images and play with stuff. But when they came to you, what did they say what what they wanted to build? What was your reaction? I mean, because I’ll be honest, when I saw the announcement, I was like,

The Harry Glorikian Show (23:52) reading this right? Is this could this be AI written? Like is this it just at at first it really took me back, which is why I texted you and I was like, what the hell is going on?

Joseph DeVivo (24:03) Right. Yep. I remember that. Well, well what what they’re doing is they’re bringing forward something called tomography. what for those who don’t know the difference, normal ultrasound device uses a reflective ultrasound. So each device can send a signal and listen. So you send a signal, it bounces off an organ tissue.

Joseph DeVivo (24:28) And then the scatter that comes back, you listen to it, and that becomes the image. And then you you process it. You actually take a lot of the noise out and you get the what you would think would be the that what would represent the it the best. But there’s also something called tomography. And that is when you take two different probes and you send a signal and then you listen to the signal. So you have one send a signal and then one listens.

Joseph DeVivo (24:55) We’ve we’ve understood tomography for a long time. We have a lot of IP in the early days of butterfly around tomography. There are companies that have created like for dense breast tomography images that like a lot of women don’t like mammography who have dense breasts because they don’t compress well. and so, you know, there’s been companies who’ve been looking at, you know, put having the you know, having the the the patient lie down on the table and having a hole in the table.

Joseph DeVivo (25:24) Allowing the breast to to go pendulent into a pool of water, and then they would do this type of imaging. So we’ve been aware of tomography for a while. but the one of the limitations, you know, with this is that you just produce so much data. And if you do so, what what they wanted to do was to put 40 devices in a circle. And if you went on Mid Journey’s website, you would find the image of the scanner, and then they have another.

Joseph DeVivo (25:53) video for those who are interested on actually how it works and and it shows the sound waves coming out. But if you imagine we have 9,000 sensors on a chip and then you put 40 of those chips in a circle, you’re dealing with three to four hundred thousand sensors that are firing simultaneously. Now and and and one of the probes will send out sound and all the other 39 will listen.

Joseph DeVivo (26:21) And then it’ll cycle to the next probe sending out sound and then the rest and it goes all the way around, firing all this sound and firing all this information. Well a normal a normal MRI can will process will create about forty teraflops worth of data. this this this tomography machine processes four hundred teraflops of data.

Joseph DeVivo (26:48) And so what they what they wanted to do is create a whole body scanner, but they wanted to use ultrasound where it wouldn’t have the mag the the magnetic issues of metal in the body. it would be very low cost, very accessible. And they and they had this vision of just creating this device that that would do this. And so literally the the first big technical challenge is could they network 40 devices? Could they get them to all talk to each other? Can they are they able to

Joseph DeVivo (27:18) discern what all this scatter and what all this noise is into an image. And it was extraordinary to watch. They they in November of 24, so a year earlier, had started buying chips. And all throughout 25 we’re buying more and more chips and building this sensor and and you know they would work on it and and they’d ask us questions and sometimes we didn’t know the answer. We would, you know, send our engineers

Joseph DeVivo (27:48) On site for weeks at a time with their engineers to figure this thing out. And and they’re just extraordinary. They’re just extraordinary brain power out there. And they came up with the ability to assemble a set of data off of all of off off of all of this scanning. And so we were blown

The Harry Glorikian Show (28:10) Yeah, I was

Joseph DeVivo (28:11) away.

The Harry Glorikian Show (28:12) I I was I that’s what I mean I was first I was thinking, okay, what can forty chips surrounding a body see that one probe in a doctor’s hand can’t? And then the other is you have to what do you pull out of that data set? That’s not a tiny data set. Like you and you need to do it within a certain amount of time. you know, you can’t take forever to do this.

The Harry Glorikian Show (28:40) Those are sort of and and I think your press release said future scanners are have even more than 40 modules, right? So you’re talking the amount of data just keeps increasing. So where’s this going? Ex explain it to I look, I I’m not I’m not an acoustical engineer, right? By any stretch of the imagination. So what do we what are we doing here and where are we trying to get to?

Joseph DeVivo (29:05) Well, I think if think about what Mid Journey does. So they they take text and they extrapolate it into a series of incredibly creative artistic images. And so they take a certain amount of data and then they represent that data in a lot of different ways. And so they have, you know, very unique capabilities and very unique capabilities that can extrapolate to something that’s very meaningful. And so

Joseph DeVivo (29:33) You know, the first step is really getting this is being able to to collect the data and capture enough of it and also do it with speed. So right now, with our current four point three chip on with the 40 sensors, it takes about twenty minutes to do a whole body scan, which is, you know, fifty percent less than what it would take to do an MRI. But we don’t think or they don’t think that, you know, someone’s gonna want to stand there for twenty minutes in a cylinder.

Joseph DeVivo (30:02) Right. And you’re being slowly lowered into a cylinder of water. then there’s a new architecture that’ll come out in twenty-eight that will get it to about eight minutes to do a scan. And then our fifth generation chip, which is called Apollo, will have twenty times the processing speed of our our our current chips. And that will be able to do this whole body scan within one minute. And we think if you can do all of this now with in one minute, then they have a product. And

Joseph DeVivo (30:32) That’s a lot of data and it’s a whole new architecture. And what it’s doing is that it’s acquiring this raw ultrasound data. And instead of filtering out the data and and trying to create it through a beamformer, it puts all the data into a Nvidia GPU using a holo scan technology that Nvidia has. And so but by by having this mass you know processing capability, it allows them to

Joseph DeVivo (31:00) A capture all this data. And I think there’s three phases to the project. It’s, you know, it’s get all these things to network so you can get a single image. But then B, take a lot of the mid-journey know-how and and take that raw data and turn it into an image that’s very meaningful. And then I think the third phase of the whole project is having AI then read the images and start telling you what it sees to help.

Joseph DeVivo (31:27) you know, filter out the false positives and then, you know, create something that’s medically necessary. So they wish to get an FDA approval. You know, they’re gonna start off as a wellness device. but they as this thing was progressing, it was an extraordinary feat to to network these things just to get a singular image. but I just think this is still early innings for them.

The Harry Glorikian Show (31:51) Yeah, I mean if you had said to me that Mid Journey was gonna go into medical, I would have been scratching my head like trying to figure out how that happened. When is the Apollo chip available? Is it already available?

The Harry Glorikian Show (32:06) If you like the conversation we have here on the show, I know you’ll like my new book, The Invisible Interface: How AI Turns Intentions into Actions and Who Wins, with a forward from Don Norman. It’s about the layer of AI forming between what you intend and what actually gets done in your business, in your healthcare, in your daily life, and who captures the value when that layer becomes invisible. It’s now available in print, ebook.

The Harry Glorikian Show (32:35) Just search for the invisible interface by Harry Glarikian, wherever you buy your books. And now back to the show.

Joseph DeVivo (32:43) No, I no the Apollo chip should be available sometime in in early twenty eight and and they should have access to it then. But you know I I do think what you’re gonna find, Harry, is you know, AI is looking for, you know, a a physical carrier. You know, right now, you know, the physical AI is next.

Joseph DeVivo (33:08) and you you hear a lot about obviously Optimus and what that’s gonna do. You know, you have the robots with with AI. But a lot of the largest tech companies are looking at ways where AI is gonna interface with the body. And so it’s not necessarily classic healthcare where you know they’re doing inpatient work and diagnosis, but there’s gonna be a lot of work from the biggest tech companies that that are out there that are looking for ways to

Joseph DeVivo (33:36) you know, to learn from, to stimulate, to to do to have individual use cases that do interface with the body. And and so I do think you’re gonna see a lot of atypical players coming into a space. And and it’s because of the power that the compute has for interpretation and for understanding just massive amounts of data. So the more data they can get, the more they can create

Joseph DeVivo (34:03) very meaningful interfaces. You know, you see the whoops, you see the Apple Watches, you see all these other ways of kind of monitoring your data. I think that whole thing is going to get much more sophisticated. And I do think that our chip is going to be core to being a portal between a lot of this AI compute capability and then access to uploading a lot of data from the human body.

The Harry Glorikian Show (34:31) you know, so I look, I’m I’m always curious about like this this journey. I mean, you’ve been doing medical devices, I’ve been in healthcare for a long time, like you’re used to going from a prototype to a product, and that’s like not for this particular product, but let’s talk about it in general, right? There’s a real distance between a demoed something and then a cleared product, as you and I both

The Harry Glorikian Show (34:59) You know, know through the pains and the scars that we’ve taken along this road over time.

The Harry Glorikian Show (35:07) What does the road look like for something like this? What are what are milestones between a demo in San Francisco and then a device? Wellness is one category, right, that you might take a journey to, but but then getting it to a physician for something is a is a different level that it needs to get to. Can you sort of give people that are listening a rough idea?

Joseph DeVivo (35:32) Well, what what I can do is I can maybe summarize what David has said in his public, you know, launch, which was, you know, in by the end of twenty I think by the end sometime in twenty seven, they’re gonna have these environments, these spas opened up where they’re gonna just be doing a lot of scanning. And they’ll and they’re gonna be learning you know, one of the things that that, you know

Joseph DeVivo (36:00) And David’s genius in building his business is he listens to his developer community. He listens to his user community. And one of the reasons why he wanted this idea out there was because he wanted feedback and wanted to understand what kind of experiences do people want and what what are they concerned with. And and so I think, you know, all through through now through the end of 27 is gonna be a lot of exper experiential work, a lot of iteration.

Joseph DeVivo (36:29) And then there’s gonna be a lot of work being done on, you know, what can be done with this data? How can they represent this data? You know, a lot of doctors initially get were pretty offended by the concept of a lot of rapid screening up front because you know, every time someone gets an out-of-course type of scan, they run to the doctor with a lot of a lot of maybe nonsense or a lot of, you know, what they call incidental omas, you know, so false positives.

Joseph DeVivo (36:58) And that does create a lot of burden for for the medical community. But if you can imagine, you know, once a week, you know, going into your health club or going in and just having a scan, you don’t even have to look at the scan. But say you get a scan, you know, every week and then you ask it three months later, you know, what’s changed in my body? What’s my BMI look like? You know, has is there anything that has changed in my organs? how is my visceral fat?

Joseph DeVivo (37:25) Yeah, I’ve been working on my legs. Have I get more muscle mass in my legs? What percentage have my legs grown in my in my heavyweight workouts or whatever? And I I think I think there’s just gonna be a lot of learning. So we’ve the first step was can we network all of these probes and all these sensors and get a cogent image? And I think the answer is yes. Now it’s how good can that image get? How and then how can we learn from that data? And then now how can that data help in inform our health journey? And I think

Joseph DeVivo (37:55) You know, once they get to, you know, twenty-eight, I think they are going to look for some regulatory clearances. They’ll start off with wellness and then they will start, you know, looking for basic clearances based upon their progress. And I do think they want this to be a full fledged body scanner by the time, you know, it comes out in twenty nine and thirty. So they have their roadmap. I th they’re gonna start with wellness because they just wanna get

Joseph DeVivo (38:21) the experiences and learn from the community. but I think they have their sights on creating a really serious device.

The Harry Glorikian Show (38:28) Yeah, I mean, I can see how like if, you know, assuming it all works, right? And it all comes together, but like Dexafit sort of stuff becomes the first area that you could go after. And if you could do it cost effectively on a regular basis, right? I can see like somebody like Lifetime, Jim’s, you know, partnering and wanting to just have that available to for a fee, of course, to their, you know, to the people that are you know.

The Harry Glorikian Show (38:56) using the gym on a regular basis. That’s a large population, not a not a small one. But

The Harry Glorikian Show (39:03) Let’s talk a little bit about physics, right? Because radiologists, I think, look at this and say, look, ultrasound doesn’t love bone, doesn’t love air, you know, lungs are full of air, you know, the bowel is air, the skull is bone. How do you guys see that working into what you’re doing? And I I mean it’s something you’re probably dealing with anyway, because of the

The Harry Glorikian Show (39:31) product that you’re making, but just just curious.

Joseph DeVivo (39:34) Well, you know, s I I sometimes y you have to think about what question am I actually asking before I I I look for the the answer. Because a normal ultrasound beam, if this is like an air-filled lung, and I have a beam that’s going into the air filled lung, the beam will stop when it hits air, right? Because sound can’t doesn’t transmute through air.

Joseph DeVivo (40:05) But what if I take it from here, from here, from here, from here, from here, from here, from here, and I’m taking a full 360-degree image. Well, I’ve never gone through the air, but I don’t need to because I have 100% of the tissue captured, right? So air is not the issue because there’s nothing in the air to diagnose. Like there’s no, you know, there’s no radiation that diagnoses something in air.

Joseph DeVivo (40:33) But radiation can travel through it. But if I’m going 360 degrees around the lung and I capture 100% of the lung tissue, then I’ve also I can see what the air pocket looks like. You know what I mean? So so air is not an issue. And that’s the same thing with the bone. Now you’re not going to be able to see bone marrow. You’re not going to go inside the bone. But because of this massive amount of data on a 360, you’ll have a complete representation.

Joseph DeVivo (41:00) of a rib cage and if there’s a broken bone and if there’s any other thing. So if we were just sending a single beam of ultrasound and and it hits a bone and stops there, then yes, but if we’re gonna go all the way around it, we’re gonna capture the exact anatomy of the bone. And we’re gonna and we’re gonna capture all the anatomy around the bowel, around the lung, and and you don’t have to go through air when you have four hundred thousand sensors

Joseph DeVivo (41:29) that are that are firing simultaneously throughout the circumference of the body. And so that’s where it’s kind of a nothing burger because the the technology itself, you know, is gonna capture all of the anatomy that you want to see.

The Harry Glorikian Show (41:45) So look, I was thinking about this. I was like, all right, how do I, how would I, you know, describe this deal to some of my f investor friends? And you got to tell me if it’s true, right? I mean, because I’m I’m pretty excited about this. And it doesn’t take a lot sometimes to get me excited about really interesting technology where I can see how it can work. But I mean, the way that you guys are going about this, I just see that Butterfly sort of wins in every scenario, right? There’s there’s a license fee or contractual, right? If

The Harry Glorikian Show (42:14) Midjourney’s vision works, you’re the imaging layer inside theoretically what could be the biggest scanner fleet on the planet. I mean, if it takes longer or if it goes in a d different direction, you’ve still collected real money. I mean, you’ve proven 40 of these chips works work in concert, and you’ve got, you know, many other partners that are starting to embed this into other products. So

The Harry Glorikian Show (42:42) You know, is that how you’re thinking about it? Am I missing something? Is there i am I not thinking about a risk properly? I I don’t know. I’m just asking the question out loud. I know that your earnings are coming up and you might be able to say something or might not be able to say something, but I’m pretty bullish on this.

Joseph DeVivo (43:00) I think it should be. And and I think I’m pretty bullish about it too, Harry. You know, we so right now Butterfly is investing a lot of money to build the point of care ultrasound business. You know, we’re putting a probe in every doctor and nurse’s hand. We have to train them, we have to develop AI, we have to develop ways to manage the data, and then we have to advertise to get new users and grow. So market development is really expensive, right?

Joseph DeVivo (43:29) You know, think of any other companies that you’re invested in or developed where you’re creating a new product for a new market and you’re changing behaviors and that’s a lot of energy. And so Butterfly has has invested all of its energy to build, be a part of building this market and accelerating this market. With the embedded part of the business, we’re not building markets. Like we’re not gonna build the market for Mid Journey or for Mendera or for Sonic Insights or for Forest.

Joseph DeVivo (43:58) You know, what we’re doing is making available six hundred patents, making available a half a billion dollars worth of investment to get to a semiconductor that’s now three, four, five generations into its development and software evolution. So we’re dramatically de-risking their ability to develop a product. But what we’re doing is not taking the risk up front to develop those markets, right? So now

Joseph DeVivo (44:24) Instead of one big shot on goal, which is point of cure ultrasound, we’re delivering for our investors multiple shots on goal. Brain computer interface companies, robotics companies, other types of imaging companies I can’t talk about yet, but all these different applications are now applications that are entirely incremental to the butterfly investor. And there’s no risk because they pay us for the software, they pay us for the chip, they pay us for the development. And all of a sudden, like Midjourney, if they say, hey,

Joseph DeVivo (44:53) This is really cool. We want to own this space. Mid Journey came to us and said, we don’t want to move forward unless we can get an exclusive license to whole body tomography with your chip. Well, that’s you know, for us to say we’re only going to do something with one partner, that comes at a cost, and that’s that \$74 million cost they were willing to pay up front to s to make the investment. So then no one else would follow them in their space with our tech. The only way someone’s going to follow them.

Joseph DeVivo (45:21) is if they come up with another foundational technology, they can get around our 600 patents and all that up front. So I think what this is is that as an investment thesis for Butterfly, it allows us a lot of shots on goal to be able to participate because you know the our relationships start off as software licensing and chip purchases.

Joseph DeVivo (45:44) But then down the road it turns into and and development fees, but down the road it turns into royalties, it turns into licensure, and it turns into chip sales. So we have this running ongoing revenue stream in all these different shots on goal now that we wouldn’t have taken as an individual butterfly. So our TAM goes from like 20 billion to 350 billion. It gets us, you know, we have shots on goal with, you know, if some things don’t work.

Joseph DeVivo (46:13) w you know, it it they don’t work, but then the the others do. And I think our investors are are able to enjoy a lot of potential upside that didn’t exist in the past for us.

The Harry Glorikian Show (46:26) So let’s jump back to like the business today. Like I think I was seeing in the Q1 press release, twenty-six and a half million in revenue. your first enterprise compass AI deal of the year. And now last time we talked, I’m and I’m going back in time, right, to twenty-three, but we talked about ghost imaging, we talked about

The Harry Glorikian Show (46:52) clinicians doing scans and never documenting our billing for them. You told me that technology wasn’t the barrier and it was workflow, which God, that that’s that’s a recurring theme for a lot of people. we’re three years on. Do you see the hospital changing?

Joseph DeVivo (47:13) Yes, I do. So three years ago I remember going to a conference and meeting with fourteen different chief medical officers from some of the largest medical systems across healthcare systems across the country. And they couldn’t spell POCUS. They had no idea what was going on. This was not on their radar. And it was very frustrating because there because I would show a list. I’d I’d

Joseph DeVivo (47:43) So-and-so medical system that’s got maybe 20 different hospitals. And I’d be like, Did you know you have 220 doctors who have butterflies and they’ve done about 30,000 scans that are not in your EMR? Are you aware of that? And they’re like, What? What’s a butterfly? What are they doing? Like at the at the C-suite level, this was not present. But today, yes, actually, those GO scans are are non-compliant, so they want to get them captured. They want that revenue capture.

Joseph DeVivo (48:13) There’s growing awareness that they can put a middleware that we have in place that can that can connect all these different point of care ultrasound devices, butterfly plus other people’s aggregate all this data, get them compliant, have them have an ROI that can complete in less than a year with a software implementation. And so it’s a big part of our business now. Our health system business is all about software first, data acquisition, aggregation.

Joseph DeVivo (48:42) Creating of centralized ways of deploying a POCUS program, a centralized way of educating people who want to learn ultrasound in certain use cases, and then having a management system with our software to help people mentor those new doctors or those new to ultrasound doctors, you know, kind of more institutionally. So, yes, we are doing a lot more deployments when after we launched, and we said this last quarter, so I can say it now.

Joseph DeVivo (49:08) That af after we launched the most recent version of our software Compass AI, our software pipeline is up. I think at the time it was up four times. I’ll update that in a couple days. But our pipeline grew 4x. And so now our teams are with the largest health systems out there talking about large deployments. And then, you know, before it was doctors buying probes, hospitals getting mad because everyone’s walking around with a probe.

Joseph DeVivo (49:37) to now it’s it it’s kind of flipping to where hospitals are putting in software and then building the programs of aggregating all of Pointy Care Ultrasound. So yeah, I think the market’s maturing and I think over the next several years it’ll be much more institutionalized than it’s been.

The Harry Glorikian Show (49:54) So when we also talk back then, you got me sort of excited about this. this picture you painted in 23 of heart failure patients going home, device on the chest, you know, beeline count streaming to a care team, right? So the diuretic gets adjusted before the readmission instead of after. And your Q1 release said a state-side deployment is expected this quarter. So that’s right now.

The Harry Glorikian Show (50:25) Can you tell us anything about that? Or is or do I have to wait for the upcoming and

Joseph DeVivo (50:29) You’re gonna have to wait forty eight hours. You’re gonna have to

Joseph DeVivo (50:32) wait forty eight hours. But what I can do though is just clarify what we said in the first quarter and what we’ve been saying, which is, you know, the the the we’re trying to build this market and in and as w other companies are. And we’re trying to get more and more people to understand that if you have an ultrasound device where the patient is, you can get quick data that can alter their care quickly at a at a low cost way.

Joseph DeVivo (50:58) And so we identified that in skilled nursing facilities, historically, there’s this revolving door. You know, they get sick, they get sent to the hospital, they get the the the they get the the diuretic modification after radiation scans, radiology scans, then they go back into the nursing home, then they get sick, and then they go back in the hospital. And about a decade ago, the government saw this revolving door. The only one who benefited was, you know, the the the fee for service providers.

Joseph DeVivo (51:25) But once they went at risk and you had to manage the whole cost of that patient, there’s a whole different consequence. And so what we’re doing is teaching at-risk providers how to teach their nurses to do ultrasound at the bedside as a part of their everyday workflow. And in order to take away all the logistics, all the complexities of software and education and hardware, we do that as a service now.

Joseph DeVivo (51:50) and so we are working on a program with a very large high at risk provider for their nurses to be able to do AI B line scans at the bedside. So then they get that scan read by a remote cardiologist, that the medic the prescription and diagnose gets modified, they get a different prescription, and they keep well our objective is to keep them healthier and stable where they are, and to reduce this revolving door cycle in and out of.

Joseph DeVivo (52:19) health system. So we’ll update you where we are in a couple days. But that is one of our ways of eating our own dog food. Saying instead of just selling you a probe and hoping that you go get trained, we put our arms around you and say, we’re going to build this program and we will take care of everything necessary to get you to that place. And if we reduce your readmissions, we’re going to share in that upside. And the and vice versa, if we don’t make any impact, then we share the downside. So

Joseph DeVivo (52:49) it’s it’s a really interesting deal. And I think what will happen is we’ll teach these at-risk providers that these nurses can do these scans. And it won’t just be for beelines. They’ll be able they’ll do cardiac echoes, they’ll do bladder scans, they’ll scan for sepsis, they’ll scan for all different types of things. And as each new AI tool comes out, it’s the it becomes easier and easier and easier for them to do it. And then this will start really changing the you know, the home base.

Joseph DeVivo (53:18) Practice in medicine. And then it’ll evolve, Harry, to where you started the conversation, where we will then have a wearable this size. And then with all the AI tools, we’ll just send it home with a patient, they’ll put it on their chest, press a button, it’ll do a heart scan, a lung scan automatically, and then they can do the diagnosis remotely. So we’re evolving towards that workflow, and we can’t be any more excited about where we are.

Joseph DeVivo (53:46) you know, with this managing home care.

The Harry Glorikian Show (53:49) So so you started bleeding into my last question. And it’s I it’s the it’s the exact same question I asked you three years ago, right? Just sort of bigger now. But 10 years from now, like you said, right, does anybody go and get an ultrasound? Or is it just like you said, imaging that happens in your home, in your doctor’s pocket, maybe in a pool of water somewhere, right? And the idea of imaging as a

The Harry Glorikian Show (54:18) Destination quietly disappears. And if that is the case.

The Harry Glorikian Show (54:26) Forget the vision for a minute, but from an execution perspective, what do you guys have to do to be at the center of that?

Joseph DeVivo (54:34) Well, I don’t think so. th think about it like a bullseye. I think the health system is at the center of the bullseye. And so you need to go and get very sophisticated subspecialty care and imaging. And you know, as much as we compete with all the big imaging companies, you know, what they do is extraordinary and what they do in cardiology and in OB and all these different use cases with the most high end sophisticated ultrasound, I just don’t see that being replaced.

Joseph DeVivo (55:03) What I see though is that if you look at the bullseye, I see the concentric rings expanding and the capability of getting closer and closer to patients where they are farther and farther away from the health system. I don’t see that as replacing the center of the bullseye. I see it expanding access to all patients. And that’s really what we’re trying to do. We’re not trying to replace awesome imaging technology. What we’re trying to do is bring imaging

Joseph DeVivo (55:32) you know, to rural going back to Africa where, you know, why should twenty thousand women d you know die in childbirth? ‘Cause you just don’t know the position of the baby. But imagine if all these rural clinics have like fifty AI tools on a butterfly and can do any scan without being classically trained, and that changes the course of their care. Think about all the rural communities, think of every patient that’s being seen by a doctor where they can intervene early and and and find ways

Joseph DeVivo (56:01) to see pathology, I just think we’re expanding the bullseye all the way out to the farthest circle, not replacing that that core.

The Harry Glorikian Show (56:13) Awesome. Well, Joe, I I I you know, I can’t wish you more goodness to happen because I think more people need care like at the point that they need it at. I know we think about the world in a certain way here in the US and maybe even Europe, but there’s a you know, a large part of the world that still needs good health care and good tools. So and goddamn, I would love to get into that, you know, mid-journey scanner one of these days and just

The Harry Glorikian Show (56:42) have it float up and down just for the experience of it. So maybe one of these days I’ll do that.

Joseph DeVivo (56:46) One of these

Joseph DeVivo (56:47) days, Harry, if we figure it out, we will get you over to San Francisco and and walk in as one of the one of the early early people being scanned.

The Harry Glorikian Show (56:56) Well, we’re living there part-time.

Joseph DeVivo (56:58) All right.

The Harry Glorikian Show (56:58) you know, my childhood home is in is in the city, so we’re we’re sort of trying to live part-time in San Francisco. j because the other thing is is from a technology perspective, I mean, just just a very different way of thinking about pushing the envelope. I mean

Joseph DeVivo (57:18) You know what

Joseph DeVivo (57:18) drives me nuts, Harry, is is on on on Saturday night I’m I’m waiting for a pizza with my son. And I get a text from an old friend of mine, my age I worked with, and he just called me to tell me that he has stage four cancer and it’s metastitized to his lungs and to his liver. And I’m just sitting there in the pizzerie and I’m starting to cry as I’m talking along. And it just breaks my heart.

Joseph DeVivo (57:45) You know, and I’m s and I’m sad and I’m this and I’m feeling this all my son’s wondering what’s wrong. And then as I’m driving home, I’m just getting angry. I’m just getting angry. Because you know how long it takes for a primary tumor to grow in your colon? It doesn’t happen in overnight. It takes years.

The Harry Glorikian Show (58:04) No. Yep.

Joseph DeVivo (58:05) It takes years. And then for it to get to grow to the extent that it actually then gets to your lymphatic system and metastasized takes years.

Joseph DeVivo (58:15) You know, I mean I mean, where’s his MRI? Where’s his scan? What is it gotta you know, and and so you there’s so many there’s so many different you know there’s so many different ailments that are asymptomatic that when you learn about them it’s too late, like f stage four pancreatic cancer, asymptomatic, you know, and so I I think we’re on the on the cusp of making a better world.

Joseph DeVivo (58:41) and I think, you know, ultrasound in in our semiconductor on chip is a way to empower people to get imaging, you know, so you’re not just going into the into the healthcare system when you’re sick, but you’re actually studying your whole life journey by knowing how your body changes and and I hope that, you know, people who are afflicted with that, the with those types of misses have many shots on gold. And you know what? Forgive me if there’s an incidental loma that goes into a GP.

Joseph DeVivo (59:10) You know, if it would save my friend or or people who have been affected by this, then then then so be it. So we’re we’re trying to craft a whole new future

The Harry Glorikian Show (59:17) I to I totally agree.

Joseph DeVivo (59:18) here.

The Harry Glorikian Show (59:19) Totally agree, Joe. And I think to myself, I’m one one of the things we I didn’t cover was like all this data, we could just make better and better sort of algorithms and models to sort of see what’s happening. Like we don’t have enough data. And if this system starts producing the right level of data at the right quality, that we can then label and sort of train the next model.

The Harry Glorikian Show (59:44) It just makes it better for future and future generations.

Joseph DeVivo (59:47) That’s exactly right. Yes.

The Harry Glorikian Show (59:50) So Joe, thank you so much for the time. It was great to have you back on the show. I look forward to staying in touch and if it’s available in San Francisco, I I’ll make myself available in San Francisco.

Joseph DeVivo (01:00:03) All right. Well thank you, my brother. I appreciate you. Appreciate the time on your show.

The Harry Glorikian Show (01:00:08) Thank you, Take care.

The Harry Glorikian Show (01:00:14) He always has a standing invitation. Three years from now, we do this again. Everyone listening, the 2023 episode is in the archive and worth your time. And the invisible interface is available wherever you buy books. Thanks for listening.

NEW BOOK

THE INVISIBLE INTERFACE

Harry’s new book, The Invisible Interface: How AI Turns Intentions into Actions and Who Wins (foreword by Don Norman), is available now in print and ebook. Get It on Amazon →