This is frankly an absurd thing for me to say. I use AI every single day. It runs a lot of my business. So this isn't a piece about staying pure and never touching the stuff. I'd be the last person who could write that one with a straight face.
I use it first and foremost to keep up with myself. I joke that I have one good idea a day, but the truth is I have countless good ideas a day. I just never had the tools to take action on them. I'm usually out over my skis in my own thinking, too many projects and thought patterns going at once. And when AI first showed up as this chat-assistant thing, it was the first time I felt like I had something that could synthesize everything I do and keep pace with me.
I got into it heavy back in March, right after I left a job and finally had the time to dive in. And what I dove into wasn't the creative power of it. Creativity has never been my problem. Managing it has. So I spent a month going through years of backlog: Notion pages, Google Drives, email chains, file systems. I built a timeline of my life. I had long conversations about lessons learned and things I'd actually done. It started as cleaning up files and turned into something closer to a mirror, a way to sit back and understand myself, and honestly to feel a little proud of myself, which I move too fast to ever really do.
At the same time I started building. My own website, then an admin panel to run it, then more and more of it, communications, CRM stuff, until the database grew into a full MCP tool that's now the brain my agents run on. My website is a monster at this point. It's not a brochure, it's the web app I run my whole business from. And building it made me realize something. This thing that would've cost hundreds of thousands of dollars a few years ago is basically a commodity now. That's a big part of why OhDavid exists in the form it does. I can hand people enormous value at a reasonable cost.
The part I actually care about is the creative work on the site, and I'm happy with it. It still took a couple months and a lot of me writing it myself. I tried, early on, to just let the AI run with the content and the design system. It didn't work. Not because the output was low quality in some luxury sense. I don't care about that. It's the quality of experience I feel when I open the site on my phone. There's a certain feeling I want it to give off, one that reflects me, and you don't get that from a prompt.
The most jarring version of this was music. I'm a musician, and I hooked AI up to Ableton and messed with one of these music generators. Watching something I'd call mine get generated that fast, in a style close to my own, rubbed me the wrong way. Because it skipped the friction. And the art lives in the friction. It lives in the struggle, the not-knowing, putting different things together that don't work until something does. Take that away and you get an output, but you don't get the thing. I haven't touched it since, and I don't plan to.
Same with scripts. I've used AI to write video scripts before, at other companies, because I felt like I had to keep up. They came out one-dimensional. Generic. And you can only fight generic for so long before you're just polishing something that was never yours.
So what does taste actually mean? For me, taste is knowing the direction. Sometimes it's as simple as "I don't like that color, I like that one." Sometimes it's a whole composition: a set of references, a feeling you're trying to evoke, knowing what the reference is and why. But underneath all of it, taste reflects values. I have a taste for things that are durable and last, not fast fashion. I have a taste for consistency. It's why I wear black jeans and a black t-shirt basically every day. Throw a jacket on it and it's fancy, flip-flops and it's the beach. I like things that are multifunctional. I like things in weird settings that stop you and make you think. That's taste, and it's mine. Not my dad's, not my government's. It usually comes from curiosity. From "oh, that's interesting, that tickled something I didn't know I had."
Here's the part people skip. I did my master's at UCLA in quantitative economics, and a big chunk of it was training machine learning recommendation models. And the AI everyone's losing their minds over right now is really just ML. Fancy ML, trained on a way wider set of data than I ever had. But a recommendation model does one thing. It predicts what you'll probably want based on what everyone sort of like you already wanted. That's an average. A very sophisticated one, but an average. I built those. So when I say AI has averages, I'm not being poetic. That's literally what's under the hood.
So the opposite of taste is that average. Tasteless is the mean of a thing. White rice. Boiled pasta. There's technically something there but you can't taste it. It's bland. We interact with products like that all the time, and not everything needs taste. Sometimes a door handle just needs to open a door. Where it gets insidious is when something is tasteless on purpose and then gets forced down your throat as the only way to be. That's the shitty version, and there's a lot of it going around.
So where's my line between AI doing the work and AI helping me do it? It's the same as my line between making a creative decision and surrendering to the process. If I ask AI to go research something, like who owns a domain, I often won't tell it where to look, because it might be more creative at finding it than I am. Let it cook. But when it's a decision about how a person is going to present themselves on their own website? I'm all over that. Because I remember how they felt in the room. I remember how they conveyed the thing that mattered to them. You can't always pull that off a transcript. That's the part I supply, and it's the part that makes the work worth anything.
This week AI helped me make some digital ads, and I'd still call them mine, because every line of text on them came from real conversations and recordings, some of it direct quotes from customers, which are theirs. The logo I made by hand. The layout was AI, and we went through a few rounds until I said I liked it. That's as much mine as anything I'd ever have hired a person to do. And I'll be the first to credit the tool. Getting those ads back was wild.
But I also pushed some shitty code this week I didn't fully test, and it caused a customer real pain. I put my name on that too. I can't blame Claude. It's literally my name on the brand. I guarantee everything I do. So I wear the good and the bad.
The line's going to keep moving, I know that. The analogy I keep coming back to: if I buy a $2,000 Gibson, it comes with great tone right out of the box. But until I do something to make it mine, it just sounds like a Gibson. Not like David. I write my own song on it, run it through some effects, change how it sounds. AI is the same. It can be an incredibly powerful tool, but until I put my own spin on it, it's just AI slop. And that word's in the water for a reason right now.
I don't know if this take ages well. But here's where I land. AI can't have taste, because it doesn't have taste buds. It knows what works and what doesn't, functionally. And even if that function is to get people to do something, it's still a function. It's not taste. My whole artistic life, I've never been motivated by controlling people. That's probably what makes me a bad marketer, if I'm honest. I think people should make their own decisions. It's why I record these voice memos instead of churning out AI blogs about how you should say fuck you to your SaaS provider and go buy a website from me. I could write that one. I'd probably make more money playing the lowest common denominator of fear. It's just not the ethos.
If we're building things for humans and not for AI, then AI won't have taste. Because how could it? Taste for what? It's ineffable. That's the whole point.
Bring yours.
