- AI with Kyle
- Posts
- AI with Kyle Daily Update 119
What’s happening in the world of AI:
Highlights
🏥 ChatGPT Health: OpenAI's Biggest Lock-In Play Yet
OpenAI just announced ChatGPT Health - a dedicated experience for managing all your health data in one place. Connect your apps, upload doctor's results, blood tests, even your genome. Use ChatGPT as a persistent health assistant.
A lot of health tech startups are having a bad day. Companies like Whoop are probably looking at this thinking: oh no. ChatGPT has 900 million users and a good chunk already use it for health or mental health questions. Now there's going to be a formal place to do that.
Kyle's take: This feels is a data lock-in play, pure and simple. Right now there's no moat keeping you on any particular AI. You can switch between Claude, Gemini, and ChatGPT with basically zero switching cost. Persistent memory helps a bit, but the other companies are adding that too.
But if you've uploaded years of medical data, nutrition tracking, workouts, test results - that's a legit moat. Moving all that to another AI would be a ballache.
Speaking of…getting this past the lawyers at OpenAI must have been a nightmare too - this is the most sensitive data possible! Any breach and OpenAI gets absolutely hammered. But if they pull it off? They'll be able to tell IPO investors they have potentially 100s of million of people's health data locked in. That's powerful.
Source: OpenAI announcement
💊 Utah Lets AI Renew Prescriptions. No Doctor Required.
Utah just became the first US state to allow AI to renew medical prescriptions. No human doctor involved.
The company Doctronic has also secured a malpractice insurance policy for their AI. That's huge. One of the protections doctors and lawyers have had against AI is that you can't sue an algorithm. Now an insurance company is willing to insure the AI's decisions. The legal liabilities (and protections against) of AI are being resolved.
So how’s it do? Well. Very well. The AI matches human doctors' treatment plans 99.2% of the time. You might think that 0.8% discrepancy is dangerous - that’s 0.8% caused by AI messing up right?
But when they reviewed those cases with human experts, the AI had made the better decision 36.1% of the time. Humans and AI tied 55% of the time. Humans were only better 9% of the time. That discrepancy is human’s making worse judgements than the AI…yeah…
Kyle's take: This is kind of game over for pharmacists doing repeat prescriptions. The AI isn't just as good as humans - it's better. And once ChatGPT Health has your complete unified medical history in one place (not fragmented across different doctors and clinics), they're in the best position to feed that into an AI for prescriptions, treatment plans, insurance... This is why that ChatGPT Health announcement is a bigger deal than it looks.
Source: Andrew Curran coverage
📉 The Pattern That's Coming For Every Industry
Saw this tweet that perfectly sums up what's happening with AI and jobs:
2021: "It can't even autocomplete a line. AI for coding is useless." 2022: "It can't even write a whole block." 2023: "It can't even pass a coding interview." 2024: "It can't even build an app." 2025: "It can't even handle complex projects." 2026: "Oh no."
Kyle's take: This pattern is going to play out in every industry at different rates. We keep moving the goalposts. "AI will be intelligent when it can beat humans at chess." Then Deep Blue wins, so we say "but it can't play Go - too many combinations to brute force." Then AlphaGo wins. "But it can't operate in the real world." Then self-driving cars. And so on.
Coding and programming are getting hit first because that's what the AI labs know. They're focusing there. But this same progression will hit every industry. Even if you're not a programmer, pay attention to what's happening to that industry now. You're watching your future.
Source: Aaron Francis tweet
🔨 I Cloned a $5 SaaS Product in 5 Minutes
Saw a guy launch a tool called "Just F'ing Cancel" - upload your bank statement, it finds all your subscriptions and gives you instant cancel links. $5 one-time fee.
I thought: that's clever. I wonder how hard it is to build?
So I built it. Took me four or five minutes to get a working prototype. Gave Claude the tweet, said "how would we build this?" It figured out the architecture - CSV parser, subscription detection, cancel link database. I said "spin it up" and it did. Dropped in a sample bank statement, it found all the subscriptions, calculated annual costs, generated cancel links. Done.
My 4 minute (working) version
Kyle's take: I'm not launching this - I'm not competing with that guy. The point is: the barrier between "idea" and "working prototype" is basically zero now. I've been to entrepreneurship events for 15 years hearing people say "I've got a billion dollar app idea, I just need to find a technical co-founder" or "I just need to raise half a million to hire developers." That excuse is gone. Done. No more hiding.
You can spin up a prototype and get it released in a day. If you've got business ideas rattling around in your head, you can just do them now.
Source: Frank DeGods tweet
🧬 I Turned My Genome Into Wall Art
Had this idea for years - take my 23andMe genome file and turn it into a big poster. Each nucleotide (adenine, cytosine, guanine, thymine) as a different coloured pixel.
Never knew how to make it. Then last night after some soju I thought: oh wait, I can just do it.
Gave Claude my 5.5 megabyte genome file. Asked it to visualise each nucleotide as a pixel. It gave me multiple colour variations - primary colours, neon, muted. I picked one, asked for higher resolution, and now I've got a 30x30 inch printable image of my genetic code.
This is zoomed in. The whole thing is a HUGE pixel fuzz
Kyle's take: Is this a business? Maybe. Hyper-personalised art from your genome. Get your genome plus your partner's genome, do some intermixing, make a piece representing your "perfect child” or whatnot. (Yes I know that’s not how DNA works, don’t @ me!)
Could be a cool gift! Again though the point is: an idea that's been bouncing around my head for years, solved after one glass of soju and 10 minutes with Claude. Your imagination is now the only limitation.
You can just build things.
Member Question from Mosin: "How long do you think most jobs have left? And do you think people getting degrees now are wasting their time?"
Kyle's response: Jobs will continue to exist. What's happening isn't wholesale replacement - it's task replacement. Every job is a bundle of 10-20 different tasks. AI will take certain tasks faster than others. Things requiring empathy, physical dexterity, or non-algorithmic thinking are safer for now. But even "safe" jobs like plumbing will lose tasks - admin, invoicing, ordering parts.
When AI takes tasks, people become more productive. Which means we need fewer people per job. That's how it insidiously affects industries. Even if the long-term future is positive, the next few decades will be messy for people directly affected - just like mining communities when those jobs disappeared.
For degrees: I'd avoid anything that locks you into one specific profession. Medical school, law school - those routes are risky if that profession gets decimated while you're studying. Something more generic that teaches critical thinking, flexibility, how to learn - that's probably the safer play. But honestly? Anyone who says they know for certain is stupidly overconfident or trying to sell you something.
Member Question from Mark are the AI Workshop Kit : "Is it a one-time fee and then the material is mine to use?"
Kyle's response: Yep. Lifetime license, no royalties. We looked at other education companies that charge royalties every time you give a workshop - how many people, how much are you charging, etc. We find that icky. We want you to have all the materials and go do well. Simple as that.
FYI we’re launching the AI Workshop Kit next Monday - details here
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.