Latest / Logosofia no Dia a Dia / Why No AR App for Infant CPR?
Transcript
- [Prompt] custom: I have mentioned in a couple of podcasts before that I really want to brush up on my first aid skill | Hosts: corn, herman [Response] Corn: Daniel sent us this one, and honestly I felt it in my bones. He's a parent in Tel Aviv, did the responsible thing, took a first aid course before the baby arrived. Six months later, he's staring at his phone at two in the morning trying to remember the infant recovery position, and what does he find? From twenty fourteen. No visuals, no interactive guidance, just a wall of text that reads like a PDF someone forgot to update. Herman: Here's the thing that makes it maddening. That phone in his hand has a LiDAR scanner. It's got a GPU that can run real-time pose estimation. It can map a room, track a hand, measure depth to the millimeter. And the most critical just-in-time knowledge we possess as a species, how to keep a human being alive for the next three minutes, is locked inside a four-hour course you took once and a PDF that looks like it was formatted for a BlackBerry. Corn: The gap is almost surreal when you say it out loud. We've got apps that can identify a plant from a single leaf, translate street signs in real time, overlay constellation maps on the night sky. But if your kid is choking? Here's a static image from the Obama administration, good luck. Herman: Daniel's frustration isn't theoretical. He and Hannah took those ad hoc courses before the birth, the ones that get organized in community centers and prenatal clinics here in Israel. Magen David Adom runs them, sometimes United Hatzalah volunteers do sessions. You sit through four hours, you practice on a little infant mannequin, you feel reasonably competent, and then you walk out the door and the clock starts ticking on everything you just learned. Corn: The forgetting curve doesn't care how motivated you are. Herman: It really doesn't. The American Heart Association has tracked this, and the numbers are brutal. CPR skill retention drops below fifty percent within six months of training. That's not six months of negligence, that's six months of normal life where your baby didn't choke and you didn't need to use the skill. By the time you do need it, half of what you learned is gone. Corn: Daniel's experience isn't a personal failure, which I think is worth naming explicitly. He's motivated, he's willing to pay for a validated app, he went actively searching. The system failed him. The system is failing basically everyone who isn't an EMT recertifying on a schedule. Herman: When he went looking specifically for an app with visuals, with actual demonstrations, something that uses the hardware we all carry, he turned up empty. I did the same search after reading his prompt. The American Red Cross First Aid app has some videos, which is better than nothing, but there's no augmented reality, no AI-driven feedback, no way to practice compressions and get told whether your depth and rate are correct. The British Red Cross app has simple animations. Magen David Adom and United Hatzalah, the two primary emergency organizations here in Israel, neither one offers a digital refresher tool with any kind of interactive component. Corn: United Hatzalah's app is genuinely impressive for what it does, it dispatches volunteers to nearby emergencies, it's saved countless lives. But it's not a training tool. It's a response tool. Different category entirely. Herman: That distinction matters because the whole model of first aid education is built on a premise that doesn't hold up. The premise is that a certification every two years is sufficient. You take the course, you get the card, you're good for twenty-four months. But the science says you started losing competence before the ink dried on the certificate. Corn: It's compliance theater dressed up as public health. Herman: That's uncomfortably accurate. And the organizations aren't malicious, they're operating within a model that prioritizes certification revenue and liability protection. An in-person course generates income and creates a clear paper trail. An app that gives real-time guidance in an emergency? That's a legal exposure nightmare for a risk-averse institution. Corn: Which brings us back to Daniel's core question. He's not just asking why the app doesn't exist. He's asking about the community of people who want this to exist, and whether anyone anywhere has actually stepped up to build it. Is there a Denmark out there that figured this out? Is there a startup that saw the gap and filled it? Or are we all just waiting for someone to get around to it? Herman: That's exactly where we need to go next, because the answer is more interesting than just "nobody's done it." Some people have. The question is why it hasn't scaled, and what it would actually take to fix this for the motivated parent at two in the morning who just needs to see, not read, what to do. Herman: Let's widen the lens, because Daniel's experience in Israel isn't unique. He ran into a wall here, but the same wall exists in London, in Chicago, in Sydney. I pulled up the digital offerings from the major first aid organizations globally, and the pattern is remarkably consistent. The American Red Cross First Aid app has videos, but they're essentially YouTube embeds inside an app wrapper. No interactivity, no feedback loop. The British Red Cross has simple animations. John Ambulance in the UK has a first aid app that's mostly text with some illustrations. None of them use the hardware that's been sitting in people's pockets for half a decade. Corn: It's not that Israel is behind the curve. The curve itself is flat. Herman: That's worth sitting with for a second, because the instinct is to say "well, Israel's a small country, maybe the market's too niche." But the American Red Cross serves over three hundred million people, and their app is essentially a reference manual with some video clips. This isn't a scale problem. It's a design philosophy problem. Corn: The design philosophy being "we've digitized the pamphlet. Herman: That's the phrase. They took the printed material and put it on a screen. Which was fine in twenty twelve. But we're now in a world where your phone can measure the depth of a room, track the motion of your hands in three dimensions, and run machine learning models locally without sending data to a server. The gap between what's possible and what's delivered is wider than it's ever been. Corn: Daniel's specific frustration is that he's not asking for a miracle. He's not saying "why hasn't someone built a holographic emergency room simulator." He just wanted visuals. Something that shows rather than tells. And even that low bar, in a country with some of the most sophisticated emergency response infrastructure in the world, wasn't met. Herman: Right, and this is where the Israeli context gets interesting. Magen David Adom is the national emergency medical service. United Hatzalah is the volunteer network, and their response times are legendary—under three minutes in urban areas, often under ninety seconds. But both organizations are optimized for response, not for preparation. Their entire operational model is built around getting a trained person to the scene fast. The idea that the person already at the scene, the parent, the bystander, might need just-in-time guidance, that's not part of the architecture. Corn: Which is a strange blind spot when you think about it. United Hatzalah's whole innovation was recognizing that the gap between an emergency and professional arrival is where lives are lost. They shrunk that gap by distributing volunteers geographically. But they didn't extend the logic one step further and ask: what if we could also distribute the knowledge? Herman: The knowledge is the thing that bridges the first thirty seconds. A United Hatzalah volunteer arriving in ninety seconds is extraordinary, but a baby who's not breathing needs intervention immediately. The parent is the first responder, whether they feel equipped or not. Corn: What Daniel's really surfaced is a category error in how we think about first aid training. We treat it as a qualification you earn, not a capability you maintain. Herman: That's it exactly. The entire ecosystem is structured around the course, the certificate, the renewal date on your calendar. It's credentialing, not competence maintenance. You wouldn't expect to retain a foreign language by taking one intensive weekend class every twenty-four months. Corn: You'd use Duolingo for ten minutes a day. Herman: And nobody has built Duolingo for first aid. That's the gap. It's not that the information is secret or expensive or hard to teach. It's that nobody has packaged it for the way human memory actually works. Short, spaced, interactive, visual, with feedback. Corn: Daniel's point about guidelines not updating frequently enough to be a technical barrier, he's right about that too. The American Heart Association updates CPR guidelines every five years. The most recent was twenty twenty, the next is expected this year. That's a glacial pace by tech standards. You could build an app today and the core protocols wouldn't change for half a decade. The maintenance burden is trivial. Herman: Which brings us back to the question of why. If the technology exists, if the demand exists, if the maintenance burden is low, and if motivated people like Daniel are literally saying "I will pay for this," why doesn't it exist? And the answer, I think, is where this gets frustrating. Herman: Let's talk about the three reasons this doesn't exist, because they're not what most people assume. The first one is liability. Imagine you're the American Red Cross and you release an app that uses AR to show someone where to place their hands for CPR. Someone uses it in an actual emergency, the outcome is bad, and the family sues. They argue the app's hand placement overlay was off by two centimeters. They argue the AI feedback said "compression depth adequate" when it wasn't. Whether those claims hold up in court almost doesn't matter. The legal exposure alone is enough to kill the project in a boardroom. Corn: The same feature that makes the app useful, the real-time guidance, is also the liability trap. A static PDF can't be sued for giving bad real-time advice because it doesn't give real-time advice. Herman: The interactivity is both the value proposition and the legal nightmare. And these organizations are not startups that can move fast and apologize later. They're institutions with reputations built over a century. One high-profile lawsuit over an AR first aid failure, and the brand damage could spill over into their core operations—blood drives, disaster response, everything. Corn: The incentive structure punishes ambition. The safer move is to keep the app as a digital pamphlet and call it a day. Herman: Which brings us to the second barrier, and it's even more uncomfortable. In-person courses are a revenue stream. The American Heart Association's CPR in Schools kit costs over six hundred dollars, and that includes mannequins, training materials, instructor guides. Organizations charge for certification, for recertification, for instructor training. If you build a effective app that lets people practice CPR on a pillow with their phone camera providing feedback on compression depth and rate, you've just offered a free or one-dollar alternative to a six-hundred-dollar product. Corn: The business model actively resists the innovation. Herman: I don't think anyone sits in a meeting and says "let's suppress the app to protect course revenue." But the organizational gravity pulls toward the thing that pays the bills. The training division has a budget, staff, facilities. The app team, if it exists at all, is a handful of people working with an external agency. The internal politics alone make it nearly impossible for the app to cannibalize the course. Corn: It's the innovator's dilemma, but with the added complication that the "disruptive" product might get you sued. Herman: And the third barrier is the one that's hardest to fix with money or policy: technical inertia. Most first aid organizations are not technology companies. They don't have in-house AR engineers or machine learning teams. When they need an app, they hire an agency. The agency builds what agencies build, which is a content delivery system. Videos, text, maybe a quiz. The RFP doesn't say "integrate LiDAR-based pose estimation for compression feedback" because the organization doesn't know to ask for that, and the agency isn't going to propose something that makes the project four times more expensive and harder to deliver. Corn: You get exactly what the system is designed to produce. A digital pamphlet. Herman: Here's the thing that makes me frustrated as someone who practiced medicine for years. The technical capability is not theoretical. Apple put LiDAR in the iPhone Pro in twenty twenty. ARKit has been available since twenty seventeen. On-device machine learning means you can run pose estimation models locally, no cloud required, which actually helps with the liability issue because you're not transmitting sensitive health data to a server. You could build an app today that uses the front-facing camera to watch someone do chest compressions on a pillow, measure the depth based on hand displacement, count the rate, and give real-time audio feedback. A little faster. Good, maintain that rhythm." That's not science fiction. That's a weekend hackathon project with the right SDKs. Corn: The barrier isn't technical feasibility. It's that the people who have the medical authority to validate the content don't have the technical capacity to build the tool, and the people with the technical capacity don't have the medical authority or the liability coverage. Herman: That's the split in a nutshell. And Daniel's prompt gets at something important here. He said he'd pay for a validated app. He's not asking for free. He's asking for something that exists and has been checked by people who know what they're doing. The willingness to pay is there for motivated users. The question is whether that's a big enough market to attract a serious player. Corn: Or whether the liability question kills it before the market question even matters. Corn: The knock-on effect go beyond just "no app exists." When people can't refresh skills, they don't just lose the technique, they lose the confidence to act at all. There's research showing bystander CPR rates are significantly lower in communities without regular training access. Not because they don't care, but because they know they've forgotten and they're terrified of making it worse. Herman: That hesitation is measured in lives. The window for effective bystander CPR is maybe four to six minutes before brain damage begins. If someone spends the first ninety seconds of that window frozen, scrolling through a mental filing cabinet trying to remember what they learned three years ago, the outcome changes. Daniel's not unusual. He's the norm. He's the motivated person who did the course and still doesn't feel ready. Multiply him by every parent, every teacher, every office worker who's ever sat through a first aid session, and you're looking at a huge population of people who want to be useful in a crisis and know they're not. Corn: Daniel asked whether organizations have sprung up to fill this. And you found a few. Herman: A handful, none of which have scaled. There's an outfit in the UK called First Aid for All, which offers free digital resources, but it's essentially a website with guides, no app, no interactivity. In the US, there's CPR Ready, a subscription service that sends video refreshers and quizzes, which is better than nothing but still essentially content delivery. Australia has Save a Life, a nonprofit that distributes free training kits with little inflatable mannequins, which is actually clever because it solves the haptic problem, but it's a physical product, not a digital tool, and it doesn't scale globally. Corn: The community-driven efforts exist but they're fragmented, underfunded, and none of them have cracked the interactive digital piece. Herman: That brings us to the private sector question Daniel raised. Why hasn't a startup swooped in? The addressable market is theoretically everyone with a pulse. But here's the brutal economics: people expect first aid information to be free. It's classified in the public mind the same way as emergency phone numbers. You don't pay for nine-one-one, why would you pay for the instructions on what to do while you wait? Corn: Even though Daniel explicitly said he would pay. Herman: Right, and I think there is a market of motivated buyers—parents, outdoor guides, coaches, people with elderly relatives. But it's not "everyone," and the conversion problem is real. A freemium model could work: basic CPR and choking free, advanced scenario packs for bleeding control, snake bites, wilderness first aid as in-app purchases. But you'd need to get to critical mass before the revenue justifies the development cost, and you'd need medical validation, which means partnering with an established organization, which brings you right back to the liability and institutional inertia problems. Corn: It's a knot. Every path out leads back to the same stuck institutions. Herman: Which is why Daniel's other suggestion, that governments could fund this, is worth taking seriously. And this is where the Denmark case study gets interesting. Denmark launched something called the Hjerteløber app, government-funded, nationwide. It shows AED locations, it has refresher videos, and it connects trained volunteers to nearby cardiac arrests. After launch, bystander CPR rates in Denmark increased by thirty percent. That's not a marginal improvement. That's a transformation. Corn: What does Israel do by comparison? Herman: The Ministry of Health mandates CPR training for certain professions—lifeguards, teachers, fitness instructors—but it doesn't fund any public refresher tool. There's no national app, no AR initiative, no digital strategy for public first aid competence. Magen David Adom and United Hatzalah do incredible response work, but the preparedness side is essentially left to the private market, which, as we've established, hasn't solved it. Corn: Denmark proves the model works. Government funding, validated content, wide distribution, measurable outcomes. The blueprint exists. Herman: It's not even expensive in the scheme of public health spending. The cost of building and maintaining an AR-enabled first aid app is a rounding error compared to what a single ICU stay costs. But public health budgets are siloed. Emergency response is one bucket, preventative health is another, digital infrastructure is a third. Nobody owns the intersection. Corn: The org chart is the enemy of the good idea. Herman: Here's where the technology has actually leapfrogged the organizational problem. Apple introduced Visual Intelligence with iOS eighteen in twenty twenty-four. You can point your phone camera at an object, a sign, a scene, and the system understands what it's looking at. Google has Circle to Search, same idea. The capability to point your phone at a bleeding wound and get step-by-step pressure and elevation instructions overlaid on the actual injury—that's technically feasible right now. Not in five years. The computer vision models exist. The AR frameworks exist. The medical protocols are well-established. Nobody has wired them together. Corn: We're sitting on a fully functional technical stack and the only missing piece is someone willing to say "I'll take the liability, I'll fund the build, and I'll get it validated. Herman: Daniel's point about the guidelines is the kicker. The American Heart Association updates CPR protocols every five years. You build the app once, you update the content every half decade. The technical maintenance is trivial. The barrier has never been the code or the guidelines or the hardware. It's the institutional will. Corn: What Daniel's really identified is a failure of imagination at the organizational level. The technology is ready, the demand is real, the public health case is solid, the Danish model proves it works. And still, the official response from most first aid organizations is a PDF and a video from the Obama era. Herman: The question becomes whether someone outside the system—a government, a well-funded nonprofit, a startup willing to eat the liability risk—eventually forces the issue. Because the status quo isn't a neutral holding pattern. Every year that passes without a tool like this, people who took the course and meant well are standing frozen in their kitchens at two in the morning, phone in hand, scrolling past bullet points while the clock runs. Herman: Where does that leave someone like Daniel, or anyone listening who's motivated and doesn't want to wait for an institution to get its act together? I think there are actually things you can do this week. Corn: That's what's missing from most conversations about this. Herman: The first one is something I've started calling the skill decay calendar. It's embarrassingly simple. You set a recurring reminder every three months, and each quarter you practice one skill. January, CPR and AED use. April, choking response for infants and adults. July, bleeding control and tourniquets. October, recovery position and seizure first aid. You spend maybe fifteen minutes, you pull up a validated video from the American Heart Association or the Red Cross on YouTube, and you physically walk through the motions. On a pillow, on a family member who's willing, on nothing but your own kitchen floor. The physical repetition is what fights the forgetting curve. Watching isn't enough. Corn: The quarterly rotation means you're not trying to review everything at once, which is what makes people give up. One skill, fifteen minutes, four times a year. Herman: And the validated video part matters. Don't just search "CPR" and click the first result. Use the official channels. The American Heart Association's YouTube channel has current, protocol-accurate demonstrations. The British Red Cross does too. These are free and they're maintained. It's not an app with AR feedback, but it's vastly better than doing nothing and hoping you remember. Corn: It solves the confidence problem you were talking about. Even just knowing you physically practiced three months ago changes whether you freeze or act. Herman: The second thing is simpler but harder for most people: demand better. If you're a parent, a teacher, a coach, email your local first aid organization. Magen David Adom here in Israel, the Red Cross wherever you are, St. Ask them directly: why don't you have an interactive refresher app? Why is your digital offering a PDF from twenty fifteen? Consumer pressure actually moves these organizations, slowly, but it moves them. They're membership-driven and donation-funded. When enough people ask the same question, it lands on someone's strategic planning document. Corn: One email feels pointless. A thousand emails from parents who took the prenatal course and six months later realized they'd forgotten everything—that's a pattern. Herman: Frame it the way Daniel framed it. Not "your app is bad." Say "I am a motivated person who wants to maintain these skills, I would pay for a validated tool, and I can't find one. What's your plan for this?" That's harder to dismiss than a complaint. Corn: The third thing is for the technically inclined listeners, and I know we have a fair number of those. Daniel's prompt mentioned AR and AI tools, and you said earlier that the core technology for an AR first aid app is basically a hackathon project with the right SDKs. How realistic is that for someone listening? Herman: ARKit from Apple has been available since twenty seventeen. It does plane detection, motion tracking, scene understanding. CoreML lets you run pose estimation models on-device. You could build a proof-of-concept that uses the front-facing camera to track hand position during CPR practice, measure compression depth by calculating hand displacement relative to a detected surface, and give audio feedback. " That's not a multi-year research project. That's a focused developer with the right documentation. Corn: The barrier, as you said, isn't the build. It's the validation and the liability. Herman: But a proof-of-concept changes the conversation. If someone builds a working demo and puts it on GitHub, it becomes a thing that exists. Organizations can point to it. Researchers can study it. Someone with funding can say "let's validate this against mannequin sensors and see if the depth tracking is accurate." Open-source projects have a way of forcing institutional hands because they remove the "can it even be done" objection. It's done. Now the question is whether you'll endorse it, improve it, or ignore it while the world moves on. Corn: The call to action for the devs listening is: be the person who makes the PDF obsolete. Build the demo. Put it out there. Let the institutions react. Herman: For everyone else, share this conversation. Send it to your local first aid organization. Say "this is what's possible, this is what Denmark did, why aren't we doing it here?" The Danish Hjerteløber app increased bystander CPR rates by thirty percent. That's not a theoretical projection. That's a measured outcome in a real country with real people. The blueprint is public. The technology is mature. The only thing missing is the organizational decision to prioritize digital refresher tools over expensive in-person recertification cycles. Corn: I think that's the thread that runs through all three of these. The skill decay calendar, the emails demanding better, the open-source demo. They're all ways of saying: we're not waiting anymore. The motivated people are here. The tools are here. The knowledge is here. Somebody needs to connect them, and if the established players won't, maybe the pressure has to come from outside. Herman: That pressure is building whether the organizations feel it yet or not. Every parent who searches for an app and finds bullet points is a future donor, a future volunteer, a future advocate who just had a quietly terrible experience with that organization's brand. Corn: The open question that's going to sit with me is this. Somebody's going to be first. Some organization, some startup, some government agency, is eventually going to launch a validated AR first aid app. And when they do, do they get a competitive advantage in public trust? Does the Danish model, thirty percent bump in bystander CPR, translate into brand loyalty and funding and relevance? Or does the first mover just become the cautionary tale after a lawsuit? Herman: I think it depends entirely on how they handle the liability question up front. If someone launches with a clear disclaimer framework, peer-reviewed validation studies, and maybe even a regulatory sandbox arrangement with a health ministry, they've built a moat. If they launch with a "move fast and break things" mentality and someone's compression depth is off by a centimeter in a courtroom exhibit—that's the kind of thing that sets the whole field back a decade. Corn: The clock on this is not neutral, because the hardware is about to change the equation entirely. We've been talking about phones because that's what people have today. But Apple Vision Pro exists. Meta's Ray-Bans are iterating fast. In three to five years, a meaningful number of people are going to have a heads-up display that can overlay instructions directly onto their field of view while their hands are free. Herman: That's the scenario where this goes from "nice to have" to "why on earth don't we have this." Imagine a parent with AR glasses. The child is choking. The glasses detect the emergency, or the parent triggers it with a voice command, and suddenly there's a visual overlay showing exactly where to position the infant, the angle for back blows, a countdown timer, haptic feedback on the rhythm. The technology to do that is not speculative. The computer vision, the pose tracking, the protocol rendering—it all exists. The ecosystem to validate it, certify it, and deploy it does not. Corn: We're heading toward a moment where the capability is so obviously present that the absence of the tool becomes indefensible. And the question is whether the first aid organizations get there first, or whether Apple or Google just build it into the operating system and bypass them entirely. Herman: That's the scenario that should be keeping someone awake at a board meeting. If the institutions don't build it, the platform companies will. And they won't ask permission. They'll ship a "Health Emergency Assistant" as a feature update, train it on publicly available protocols, and disclaim liability in a forty-page terms of service nobody reads. It won't be validated by the American Heart Association, but it'll be on a billion devices by Tuesday. Corn: At that point, the validation question becomes secondary to the distribution question. People will use what's in front of them. Herman: Which is why the window for the established organizations to lead on this is closing. They have the medical authority now. They won't have a monopoly on distribution for much longer. Corn: For Daniel, and for everyone listening who's been nodding along, here's where we land. If you've found a good first aid refresher app or resource, something that actually works, something with visuals and interactivity and not just bullet points, email us. prompts at my weird prompts dot com. We'll compile a list for the show notes. If enough people surface things we haven't found, we'll do a follow-up. Herman: If you're sitting there thinking "I could build a prototype of this," build it. Put it somewhere we can find it. The gap between what's possible and what's available is the widest I've seen in any domain that actually matters for human survival. That's either an indictment or an invitation, depending on what you do next. Corn: Now: Hilbert's daily fun fact. Hilbert: In nineteen twelve, a microbiologist in Bishkek documented a strain of smoked horse-milk cheese whose rind contained a bacterium that produced trace amounts of trimethylamine, the same compound responsible for the odor of rotting fish. Herman:...I have so many follow-up questions and I'm going to suppress all of them. Corn: That's the right instinct. This has been My Weird Prompts. Our producer is Hilbert Flumingtop. If you enjoyed this episode, leave us a review wherever you listen, it helps other people find the show. We're back next week.