Just pick an AI model and start already

They are all amazing, just pick one, dive in, and experiment

I’ve been commuting for a new job for the past while. Which means, I’m back listening to podcasts. And me being me, I’m listening to mostly marketing podcasts, AI podcasts, marketing with AI podcasts, and tech news podcasts. I hear a lot of reviews every time a new model comes out. Bench scores—I can’t keep them straight and I tune the scores out half the time—costs—this I perk up for because I always try to maximize my usage—and what people say the new model is best at.

But I’m an edge case. For pretty much every regular AI user out there it doesn’t matter. It just doesn’t.

TL;DR: All the major AI models—Claude, Gemini, ChatGPT—are genuinely excellent. You are not losing out by picking one over another. Whichever one your company pays for or you already like using is fine. What actually matters is committing to a single tool, paying for it, and spending real time playing with it until you find the things it can do that you never thought to ask for.

On the AI Daily Brief recently, the host read a comment by someone on X which was essentially what I’ve been saying for over a year—all the models are really, really good. You’re not losing out picking one over the other. Don’t get hung up on what the “experts” say, just start using the tools.

Does it matter which AI model you use?

No, not really. Or not in a way that matters to 99% of people.

Yes, they all have their strengths and they all have their weaknesses, and I am really enjoying using Claude Pro, but Antigravity still did a really great redesign of my website. I’ve been using Claude heavily at work, but also still using both Gemini and Claude at home. Both great tools, both have their place in my toolkit. Gemini can transcribe audio and make images, Claude can’t. Gemini is interwoven with Google Cloud Platform and Firebase, so it’s easier to work with those in Gemini.

Claude?

A much better toolset ecosystem than Gemini—even with the newest model releases. It’s much easier to set up tasks that will just go off and run in Claude than Gemini (though Gemini Spark is getting pretty close). It’s also way easier to use Claude from my iPhone or iPad than Gemini. Sometimes I start things in Claude on a mobile device forgetting that it’s doing all this on my Mac and am amazed at all the things that are just done when I sit down to my Mac again.

Am I leaning way more towards Claude right now than I was for Gemini?

Yeah.

Here’s the thing, if your company pays for Google Workspace and an extra AI lift for Gemini, you’re fine. Your company pays for ChatGPT? That’s fine. Even if the company says you can only use ChatGPT in certain ways and limits things, or pays for Claude but have turned off some parts of it or limit whatever. It’s fine. All of these models do amazing, mind-blowing things.

Does Opus vs. Sonnet vs. Sol vs. Terra vs. whatever actually matter?

The folks who get obsessive about whether they’re using Opus 5 or Sonnet 5. Do they use medium or high or extra high effort. Or blow the doors off the world with Fable 5 on Claude or Sol on ChatGPT, are the fringe. It really doesn’t matter for 99% of people. Someone commented on one of the AI podcasts that getting that into the weeds about models is like debating what your favorite color is or what your favorite Pokemon is. All the colors are nice (I can’t speak for Pokemons). It’s all kind of silly, really.

Why playing with AI matters more than optimizing it

We can do so many amazing things. So many, many amazing things that the only advice that matters is—pick one, pay for it, and then dive into one tool and learn it.

Goof around. Try experiments. “Gee I wonder if it can take this hand drawn sketch of a workflow and turn it into a workflow diagram”—by the way Claude could do that. See if it can take scribbled note of three columns of numbers I’ve written down and tally them and sum them—yep Gemini did great at that.

Just throw things at the tool and see what happens. Approach all of them with childlike wonder and ask: “Gee I wonder if I can…”

And try.

Try fun stuff. Make yourself a new website, make yourself a new app. Make the thing you’d love to have, but either no one makes or isn’t quite right. Sit down with your kids and make a game—even if it’s completely silly. Just do a thing because it’s fun. That sense of play, just experimenting with really no direction, that’s where the magic is. That’s how you discover new, amazing things.

How using AI for fun makes using AI for work more productive

Yes do the work stuff. Do the research, do the summarization, do the email triage and document analysis and the “please make sense of this Excel spreadsheet for me because I can’t make heads or tails of it” stuff.

Of course do those things. But do all the other stuff too because that is where you get to the point of: “Hey, I don’t want to do this workflow anymore, this process is really dumb. I wonder if I could do it completely differently…”

Well maybe now you can do it. Maybe you can change the entire workflow. But you’re not going to get there or know how to be ready for AI minions doing your bidding, if you don’t start trying all the things. You’re not going to be able to imagine it’s even possible.

Where are you in the AI adoption curve?

We’re in a whole new world. It’s like we suddenly discovered how to make steel or any of the other massive technological breakthroughs of humanity. We’re still figuring out forges, and blowers, and how to get the right mix of iron and carbon to make great steel. We’re still at a point where we’re trying to imagine the world as a new place with new technologies.

We’ll get there, but to get there we all have to start goofing around with the tools to find that next cool thing to do.

And it doesn’t matter where you start or which tool you use, what matters is to just start.