I’m as Boring With My AI as I Am With My Investments

When Grokbot hit the scene, I was enamored.

I could see the vision right away. And here’s the rare part: I wasn’t disappointed. I dove straight in, knowing exactly what I wanted to do with it, and I’m really happy with the tool. (I wrote about the first job I gave it in I Gave a Bot the Job I Was Bad At.)

It also pulled me back to Cursor, which I hadn’t used in a while. Cursor impressed me. Grokbot impressed me more.

Then everybody asked about the next thing

Now there’s OpenAI’s Dots. And Anthropic’s Duet is coming eventually, probably very soon, from what I hear. By the time you read this there will be another one.

Have I tried Dots? Nope. I haven’t fucked with it.

I know there are other autonomous agents out there. I’ve even tried building my own, and it’s hard. Not straightforward at all. I also tried Hermes and did not like it. Pain in the butt.

And before you ask: yes, I know this is the opposite of what a guy who builds websites and workflows for a living is “supposed” to do. I’m supposed to be testing everything, all the time. I’m just not wired for it, and chasing every launch wears me out.

So I stick to my tools.

I’m boring with my AI on purpose

Here’s the best way I can explain it: I’m as boring with my AI as I am with my investments, for better or worse.

Am I missing out on money by being boring with investing? Sure. Am I missing the latest innovations by being boring with AI? Yes.

But shiny object syndrome is a thing, and it’s exhausting. I’d rather be a little behind and sane than current and fried.

Boring has its upsides. Boring means I get to spend that energy on the actual work instead of on the next evaluation.

What the boring stack looks like

I’ve been on the Claude train for a couple of years now. Anthropic is really good at making tools for knowledge workers, and the toolset capabilities are great. That’s where my work lives.

When I need to create media, I offload it to platform models like Oxen or Atlas. (I compared a bunch of those in Six AI Media Platforms Compared.) I don’t need Claude to do all the things.

I also don’t need to test every other open-weight model. I’m sure they’re awesome. I have colleagues who use them with great success. I’m just telling you what I do.

Pick a couple, drive them into the ground

That’s the whole method. Pick a couple of things that work really well and use them until you know where they bend.

It keeps me sane. It’s also how you really get to know a product’s capabilities, which is the only way to figure out if you actually want to swap it out. You can’t judge a tool you’ve poked at for a weekend. (The day-to-day version is in My WordPress Development Workflow, 2026.)

How I decide if something earns a trial

If you run a business and your inbox is full of “did you try this?”, here’s the filter I use, built from my own habits:

• Can I name my first job for it? With Grokbot I knew exactly what I wanted to do on day one. If I can’t name the job, I’m just window shopping.
• Does it fix something I’m actually stuck on? Cool demos don’t count. A real problem in my week does.
• Is it a swap or an add? I try to swap a tool out, not pile another one on. More tools means more tabs, more logins, more things to babysit.
• Is the pain worth it? Building my own agent was hard, and Hermes was a pain in the butt. If setup feels like a second job, I bail.
• Does a specialist already do this part? That’s why media goes to Oxen or Atlas instead of me forcing one tool to do everything.

If it doesn’t clear those, the answer is “not right now.” That’s allowed. Nobody gives out medals for trying everything.

And if you want a second set of eyes on which tools belong in your stack, that’s the kind of thing we sort out in AI consulting.

No, no, and no

People always ask: “Did you try this? That? This other thing?”

No. No. And no. Thanks.

Thank you.

So what’s in your boring stack? Which one or two tools would you drive into the ground?