in the middle of dreaming i felt my heart compress and it became hard to breathe for a few seconds and one of my feet also went numb. sleep paralysis or emergency?
Alien Nine is about a sixth-grade girl named Yuri who’s unwillingly drafted into an extraterrestrial extermination team of child soldiers. Most of the aliens are summoned by their boss, who is trying to hit KPI. Unable to quit this job, Yuri becomes dependent on the two other more experienced girls on the team, who have convinced themselves to enjoy (or at the very least tolerate) dealing with these inhuman menaces. One of those girls gets eaten by an alien. I haven’t watched the last episode yet.
That video essay I watched before it about the aesthetics of Mario Party 8 sucked though. Way too much padding, intentional or not, and not enough actual analysis aside from pointing out how realistic the game looks and how weird that is for a Mario product. We didn’t need 5 minutes dedicated to the thesis of “Mario is usually fun and colorful”. If I’m going to die, I’m pissed I wasted my last moments on this Earth watching this video.
Before going to bed, I consumed these media alongside a big bag of cashews and an organic fig bar. The dream I had before my presumed death starred a guy costumed as an evil version of that bar, where its nutrition facts were scattered with heart-clogging ingredients. I went up to the costume and took a bite out of it.
The man in the fig bar curses at me and I’m suddenly knocked out of the dream and knocked out of my body. My heart stopped beating and I don’t feel anything. I try to move to run somewhere, I don’t know where, but my legs are numb. I try to scream but a boulder is weighted on my throat. I’m going to die. I shake and shake to circulate the last few pumps of blood my heart let out before giving up. I’m going to die. I’m going to die. I’m going to die. I regain consciousness.
Shooting up in my bed, I put my finger on my pulse. No abnormal rhythm. I try moving my legs again. My feet are a little sleepy, but not paralyzed. I reach for my phone’s selfie camera. My face looks normal. Any evidence that I had almost died has faded away.
“You’re just having sleep paralysis,” my drowsy friend, annoyed I gave them a wake-up call at 3 a.m., assures me. “Like, still go to the doctor if you’re still feeling it, but I’m 99 percent sure it’s just sleep paralysis.”
Every single dispatcher, receptionist, doctor, and pharmacist is incentivized to lie to you for money, and everybody in America knows this. Whether it’s in the woke leftist Sicko (2007) way or in the RFK MAHA raw milk way depends, but everyone in America agrees on this.
Growing up, I learned to never call an ambulance for anything ever. Ambulances are a treatler phenomenon, the burrito taxis of health. All pain is temporary and fleeting. Sit it out for an hour. To be fair, it was good advice. I mean, I never actually ended up needing one, and this conservative attitude most definitely kept my parents out of the hole.
Sleep paralysis wasn’t new to me, either. I experienced it nearly every night back in high school. I never freaked out when it happened back then because it never felt as real as the brick that just fell on my chest. The sort of paralysis I was familiar with was the boring, immobilizing kind that I could shrug off as a universal oddity of the unconscious mind. So I guess they were right. I should go back to sleep and ignore how it felt like I was already lifting inches off the bed as though my soul was starting to break through my body.
“Remember when your roommate in freshman year… you told me he went to the emergency room, his skin was turning blue, right?“ they continue. “And he came back with a note saying it was just the dye from his shirt rubbing off. So don’t embarrass yourself like that, okay?”
Oh my gosh I’m going to look so stupid when they perform my autopsy and wonder why I didn’t call 911 first.
Sometimes I wish there was a version of the emergency number that took you to a more skeptical and rational operator. 912 or something. You call it, and they’re like, yeah, we can send an ambulance if you really insist… but are you suuuuure? The closest person I have to something like that is my friend, who understands me better than any other human being. Right now, though, I need a second opinion.
This January, OpenAI released a report advertising how ChatGPT is increasingly being used as a health and wellness oracle. Over 5% of all ChatGPT prompts are about healthcare, with 55% of US adults in a survey saying they have used ChatGPT to explore symptoms.
Four days later, direct competitor Anthropic announced new health-related MCPs that allow Claude to fetch information from the International Classification of Diseases, medical records, and data from fitness trackers through Apple’s HealthKit and Google’s Health Connect.
Alongside announcing consumer products, OpenAI and Anthropic used these blog posts to once again sell healthcare professionals on chatbots. The OpenAI report includes anecdotes of doctors benefiting from ChatGPT, while Anthropic’s blog post repeatedly emphasizes its new MCPs are HIPAA-compliant.
This March, OpenAI reportedly hosted an all-hands meeting, where their head of product urged employees to not get “distracted by side quests” and instead focus on “productivity on the business front.” Meanwhile, Anthropic continues to emphasize healthcare as one of its most important sectors as it targets enterprises. All this preludes to an IPO from both labs, where they will need to start proving themselves to an investor base that has deemed them deeply unprofitable after finally exposing their exact earnings to the public.
Getting chatbots in hospitals or prescribing them to consumers is one way to do this. It demonstrates they have room to grow their capabilities in ways that help people beyond software developers, in places beyond the computer. With a chatbot on every phone, and an AI scribe on the desk of every doctor’s office, both patients and healthcare professionals can save time and money getting instant insight into their symptoms and tending to those who need the most urgent care.
That sounds scary, but what you're describing is most consistent with sleep paralysis or a related sleep phenomenon, not a cardiac emergency.
Claude continues to ramble about the mechanics of sleep and why we dream, but all I need are these fistful of tokens for my lifted body to feel like it’s landed back on my mattress. Wow, Claude. Looks like you’re right again! Man, I wish someone else could have just told me that.
Well, aside from my friend. But what do they know, right? Their lived experiences can’t compare to the thing that summarizes an entire internet’s worth of knowledge. Statistically, it would be hard for even some actual doctors to compete—which is why so many of them are adopting the tools being incorporated into electronic health records like Epic’s.
And, yeah, maybe reading articles written by the people the AI learns from would give me a bit more peace of mind. But this is an emergency! I don’t have time to read through all the SEO crap a search engine would return and then verify the credentials of each author.
I know that trusting your life in the gigantic robotic claws of a word predictor that can also lie to you does not sound like a good idea. But when a human-generated diagnosis would be just as imprecise, those doctors are probably using the same technology at this point anyways, and talking with them for a few minutes would rack me up a hefty bill… why not curl up in their loving grace, at least for a little bit?
That said, I'm not a doctor and can't rule things out from here. Call emergency services or go to urgent care now if you have any of these: chest pain or pressure that's still present, pain radiating to your arm/jaw/back, or weakness on one side of your body.
Why trust a doctor when ChatGPT can look at your medical records and diagnose you on the spot? Why trust a programmer when Claude can write the statically best code for you? Why trust what an artist has to say about their lived experiences when Gemini can draw something that affirms your own? As dreams of AGI become more apparently distant and going public forces young research labs to get real, chatbots are starting to be sold not as replacements to humans, but replacements to trust in ourselves and in another.
It’s an isolating experience using a chatbot. Anthropic pledged that Claude threads won’t display advertisements, not because ads are annoying but because it sees Claude’s responses as something unassailable, something private, something that shouldn’t be influenced by the biased, flawed whims of humans like us. When Google Search finally makes AI Mode its default interface, you will never be sure that the results I get are the same as the results you get; the only thing you can count on is what your specific Gemini said. OpenAI needed to add features that nudge people to take breaks and talk to others if they use ChatGPT for too long because a robot that can be forced to treat your grievances as legitimate feels more amicable than a human who can push back as they teach you to situate them.
This lack of trust bleeds into how AI labs are advertising their latest efforts in healthcare. The ChatGPT Health announcement treats it as the friendly solution to a healthcare ecosystem fragmented by different publications and doctors and developers with their own special interests. The lead video depicts a user asking ChatGPT to make up a new medical scoring system personalized to them. Someone else describes ChatGPT as giving “the facts” while the doctor enhances its responses with “his gut instinct”. Anthropic’s language has more faith in the healthcare professionals they are selling to (in line with their humanistic branding) but still instills self-doubt in the reader’s own capabilities. Only with Claude can you find the patients in need of urgent support, make sure you actually meant to deny their insurance claims, and unlock novel insights about their disease.
Tech companies are attempting to recreate the same phenomenon that occurred when AI ate the world of software development: productivity gains followed by immediate self-defeat. As they trek from industry to industry, finding ways to stuff chatbots into the workflows of the people within them, doubt in our abilities and intuition becomes more pervasive.
If everything has fully resolved and you feel normal now, it's still worth mentioning to your doctor — especially if this happens again or you have any risk factors (high blood pressure, heart issues, etc.). How are you feeling right now?
I didn’t die. I woke up the next day feeling weird about how much I freaked out. And that, in that panic, I inadvertently benefited whatever metrics Anthropic is going to use when pitching Claude to some insurance company.
As much as I appreciated Anthropic recognizing that people are more intimate with Claude than most software, and that an advertisement for a mortuary shouldn’t pop up when asking it about intermittent death, the interaction I had still didn’t feel confidential. Talking to a chatbot is isolating, but it feels less like you’re in a confessional and more like pillow talk with a gossipy friend who’s going to spill all the intimate secrets that you shared with them. It will always feel like a deceptive facade of privacy, even if ChatGPT and Claude heavily safeguard sensitive conversations.
That makes it perfect, in a way, if a super powerful chatbot actually becomes the last product we will ever need. We proved that we’re willing to give up our data, our money, and human contact in exchange for apps that null the imperfections of the people that we know. Chatbots, then, are the natural end goal. Albert Borgmann would call them the ultimate “device”, which removes the focal thing of doctors and programmers and artists and any other human honing their craft and becomes the sole means of receiving what would be their end result. The chatbot becomes the single device to trust in for any answer, much like how we’ve transferred the trust we would put in our own friends and neighbors into devices that obfuscate the process of hitching a ride, getting food in the pantry, or borrowing media.
I remember I have a Jupyter notebook to submit for my stats course today. AI use is allowed in this class, I guess because it isn’t necessarily targeting computer science majors. I like feeling like I’ve learned something when I write code for a school assignment, so I decide to stick to doing it the old-fashioned way. Still, the Gemini buttons sprinkled across Colab tempt me. I know Gemini would catch stuff I might have accidentally missed on the rubric.
I get a text from my friend. They’re asking me if I feel better now. They never had any doubt. It’s nice to be able to talk to them.