Superintelligence Hasn't Reached the ICU Yet — Why Healthcare Apps Still Matter

I came across Pieter Levels' recent blog post about superintelligence racing past us, and it's the kind of provocative take that sticks in your head. He paints a picture of AI equalizing everyone, turning app development into a slop factory where differentiation is evaporating. The only winners? The superintelligence itself and the companies serving it. It's a compelling, if unsettling, vision — and for anyone building traditional consumer apps, it might feel like the ground is shifting.

But here's the thing: our data at PainSignal shows that the app layer is very much alive in places Levels doesn't look. While he's right that consumer AI slop is real, he misses the entire world of regulated, high-stakes industries where pain points are severe, unsolved, and screaming for domain-specific solutions. Take healthcare. We track 703 problems there, with an average severity score above 4/5. These aren't "build a better to-do list" problems. They're things like medication errors from IV pump malfunctions (severity 5/5), nurse burnout from understaffing (severity 5/5), and no real-time escalations for critical vitals (severity 5/5). People are literally dying because the right software doesn't exist, and no amount of generic AI can fix that without deep integration, regulation, and domain expertise.

Levels says non-technical people are now building the same or better things than technical people. That might be true when you're cloning a habit tracker, but in healthcare, the barrier isn't just coding speed. It's understanding HIPAA compliance, interfacing with ancient EHR systems, and designing workflows that work in a chaotic clinical environment. Our pain data shows that many top opportunities require complex architecture and hardware integration. For example, "LineGuard RX" — a real-time IV line monitoring system to prevent med errors — isn't something you prompt Claude Code to whip up in an afternoon. It needs sensors, FDA approval, and integration with existing hospital infrastructure. That's not slop; it's a defensible business.

This is where Levels' point about the physical world starts to resonate, but he only talks about getting your hands dirty with tools and materials. He doesn't go far enough. The real moat isn't just manual labor; it's regulated labor. In healthcare, the problems with the highest severity and willingness to pay are ones where you can't bypass the system with a clever LLM. You have to work within it, and that takes time, money, and specialized team members. For a vibe coder looking at the landscape, that might sound discouraging, but it's actually a massive signal: While everyone else races to the bottom building AI-generated consumer slop, the real untapped value is in the muck of regulation and hardware. Build something that requires blood, sweat, and compliance, and you're not just copying a carrot — you're planting a garden.

Our data also challenges the idea that the app layer is decimated. We see explicit willingness to pay for solutions that address these severe healthcare pain points. From hospital inventory management systems to nurse scheduling burnout, the common thread is that professionals aren't asking for more generic AI; they're asking for apps that understand their specific, messy reality. When Levels says he doesn't even use apps anymore and just asks Claude Code to do things, that works for personal health data analysis, but what about a ICU nurse who needs to coordinate care across ten patients, three doctors, and a dozen devices? The AI assistant isn't replacing the orchestration layer; it's demanding that it be smarter and more integrated.

Two stats from our data stand out. First, across those 703 healthcare problems, the average willingness-to-pay score is high enough to build sustainable businesses around. Second, the problems rated severity 5/5 — the life-or-death ones — almost all require blending software, hardware, and regulation. That's not an accident. The market is telling us where the durable opportunities are.

So, what's the takeaway for builders? If you're in the consumer app game, Levels' warning is worth heeding. Differentiation through speed is a fragile strategy. But if you're willing to step into a regulated, physical-world industry like healthcare, the game changes entirely. Pick a verified high-severity pain point from our dataset, one where the solution can't be generated in three minutes. Build the hardware sensor or the EHR integration. Learn the compliance landscape. That's the work that won't be automated tomorrow, and it's where the real value lies.

Levels ends by saying he has no clue where things are going. I'd argue the direction is clear: The superintelligence has indeed passed by the app stores, but it hasn't reached the ICU yet. For those of us willing to build in the real world, the carrot isn't running away — it's just hidden in the hardest places.