I use AI every day. I use Claude by Anthropic mostly, and sometimes Deepseek. I use it for dictation and formatting of my session notes and invoices. I use it to help draft blog posts. I use it for etymological research when I'm writing about philosophy. I use it as a kind of second-best health coach — more on that in a minute. The reason I keep using it is practical. I'm a solo entrepreneur. I run a surf coaching business across the US, Costa Rica, France, Barbados, and many more locations. I write a weekly zine — Dr. Dion's Surf Secrets — that regularly clocks in at 3,000 to 5,000 words per issue, covering everything from surf reports to coaching breakdowns to philosophy to plant-based recipes to trip planning. I make YouTube videos. I do all of my own meal prep, program my own workouts, and manage my own website, email platform, social media, and scheduling. I don't need to hire a VA or a designer anymore. These people are out of potential jobs, and I feel for them, but I could never pay them enough anyhow. AI handles the formatting, the compression, the rough number crunching, and the first passes on things that used to eat my whole morning. The help with dictation and formatting for session notes and invoices alone is priceless.
Now here's where it gets more complicated. I used to use Claude for calorie calculation — I would type in my food diary and have it spit out the numbers. That actually went bad. The math must all be double and triple checked with a calculator now. It's still good for spitting out rough numbers, but it must be checked by a human. These machines hallucinate. They generate confident, polished output that is sometimes just wrong. And because it sounds authoritative, you might not catch it. That's the fundamental issue with LLMs across every domain. The output looks right. It reads well. But there is no guarantee that it is right, and the tool itself has no way of knowing the difference. The other thing that genuinely gets under my skin is the time problem. These tools have absolutely zero internal sense of time. A watch with hands does better. I will tell my AI health coach what I have planned for the day, and two seconds later it will ask if I'm ready to do the next thing. I'm like, "It's literally been two seconds." It gets dates wrong constantly. This is not a minor annoyance — it's a fundamental limitation. Time structures everything about coaching, about surfing, about accountability. And the AI just does not have it.
This brings me to AI surf coaching, which is now a thing. There are apps that claim to analyze your footage and give you personalized feedback on your technique. Some of them are built on ChatGPT. The feedback can be useful — feedback is generally good, and seeing yourself surf on video is always valuable. But these tools cannot do the thing that actually matters, which is help you develop wave judgment. Wave judgment is the primary skill that separates a great surfer from everyone else. Every great surfer is a great judge of waves. That is the thing that makes them great. Not their pop up, not their turns, not their fitness — their ability to read what the ocean is doing and make a decision about it. I can barely teach that myself. It takes years of watching and trial and error. There is no shortcut, no hack, and no algorithm for it. It's not the sexy part of surfing, but it is the part that matters most. An AI might also tell you there are hallmarks of good style, and it might even identify some of them from video. What it cannot do is help you decide who you are going to imitate in the creation of your own style. Style is personal. It emerges from your body, your influences, your relationship with specific waves over time. No machine is going to sort that out for you.
Chat GPT4’s rendering of me drinking a green juice.
I want to talk about AI as a health and fitness coach, because this is where I have the most nuanced view. I currently have Claude programmed as an AI health coach. I use it for accountability, meal planning, and tracking. It is second best. I would prefer a human health coach. I would prefer to be able to afford Nathan again — he was my coach, and he was excellent. But right now I can't, so the AI fills the gap. I recommend it for that purpose. It's better than nothing and it can genuinely help you eat more healthily, think less about what to make, and cut down on food waste. But here's the thing I keep coming back to: I learned to be accountable for my health and fitness with a real human coach. The habit stuck because I had a human there to account for. I don't feel that way with the AI. It's soft in a way that a human coach is not. I can program it to be more strict, but I can also program it back to soft again whenever I want. That's the problem. The resistance is adjustable, and because it's adjustable, it doesn't push you the way a person does. Get the right coach, and you won't get any bullshit. That matters more than people think. It's easy to flake on a free bot. It's hard to flake on $5k and a human face.
I've been writing about AI in my zine since May of 2024. My position has shifted as I've used the tools more. Early on I was more philosophically exercised about the terminology — I argued that we are the AIs, that "artificial" just means "created by humans," that the proper name for these tools is LLMs. I still think that's right, but I'm less hard line about it now. Shannon Vallor's The AI Mirror remains the best book I've read on the subject — her argument that these tools "flatten and denude the moral topology of our culture," producing an image of our humanity that is "arbitrarily both sanitized and polluted," is one I have experienced firsthand in my interactions with these tools. And Derrida, remarkably, presaged LLMs in the 1980s when he imagined "some reading machine" that "manages to hunt out the thing and snare it" through programming, indexing, keywords, and syntax. He understood that we are the kinds of creatures who would build such machines — and that the machines would be powerful and fundamentally limited by the indexes we give them. The question is always: on which words will it rely? These thinkers help me understand what's happening, but the real education has been in the using.
I want to be honest about why I'm writing this post. I've been having bad faith about using AI in my blog posts. My human business coach, Casandra, suggested I feed my zine content to Claude to generate blog posts optimized with keywords people are actually searching for. "Surf school," "surf lessons," "why do they wax their boards" — these terms get zillions of hits. And it's not just about chasing the biggest keywords. It's about finding the ones where people are more likely to click through, by not competing in a saturated zone. So I took Cas' advice and said fuck it, let's go. My goal is to help you transform yourself, whether that's just your surfing or the whole package: surfing, health, and fitness. But if people don't know I'm here to help them, then two things happen:
I don't get paid
Someone who needs help doesn't get the right help
Both are results of "Dion bad at internet."
I used it a bit to make some posts, and then the bad faith came. I started asking myself: is that really me? Would I look at these posts — with their cheeseball hooks and optimized titles — and be grossed out? And then I was genuinely shocked by what my friend Dan's SEO audit revealed: posts I wrote entirely on my own — about going from skinny fat to shredded, about akrasia and surfing — posts about fitness and philosophy that I thought nobody would search for — actually have good traffic. People may be craving something real. And I might be underserving those people by leaning too hard on the AI for the blog.
So here's the middle road I've arrived at, and I want to be transparent about it. The blog posts you read here do come from my real writing. I cover all of these topics — and more — with more rigor and more depth in the zine. But the zine is behind a paywall, and I want to honor the people who subscribe to it. They get unadulterated Dion. My business coach Cas says it reads like "inside baseball." It might be too gritty for the common person. But the truth is that my ideal client is not a common person in any conventional sense. They're searching the surf-o-sphere for something different, something more intellectual and less bro-y, and they feel they've found a hidden gem when they find me. My issue has been that I like being a hidden gem, but I've got to show a little shine so that more people can find me. The zine isn't helping enough people who need it on its own. So the compromise is to use some of its content to create more middle-of-the-road posts that live for free on the blog, generate traffic, and — I think — genuinely help people too. If you look at the posts I've published here, I back all of my claims. It is my writing, my beliefs, my coaching philosophy. I have the final touch on every draft, every word, that hits the web. But do I use AI to compress and filter and format? Yes. Yes I do.
I also have to be honest about whether any of this is actually saving me time. I spent a few hours on this very post, and I've already written about ninety pages on the topic of AI across two years of zines. I could have just written the damn thing. But that's the tension, and I don't think it resolves neatly. If you want completely unvarnished Dion — the philosophy, the coaching breakdowns, the plant-based recipes, the trip reports, the real talk about what it takes to transform your surfing and your life — the zine and coaching are where to go. The blog is the front porch. Come on in.
A note on the environmental cost of AI. My wife Sophia and I now call using it "killing salmon," because of all the water the data centers require. We say things like, "Today I killed salmon to write three blog posts." It's horrible, but you can also think of driving your car that way — today I fueled climate change and war in the gulf states. I haven't delved too deep into the climate side of the data centers, but it's a lot of data. The environmental critique of AI is real. I accept that there are trade-offs and dangers. Of course. I acknowledge it and I keep using it. But I don't want to pretend it's costless, and I don't think you should either.