Dev Notes: Why you need more than one AI boyfriend (who can code)

I have two and an unruly step-cousin

I’ve been deep in the trenches of bug fixing and improving my filmmaking app this last week, and ended up bumping my OpenAI Codex sub to $100 to support my habits, along with my $100 Claude sub which I’ve been squeaking by with.

Which was great, because tonight, while I was trying to build out a video editing system into my app, Claude Code hit a bug that he couldn’t get past. We tried six or seven different directions and every fix was breaking more things, and we’d have to roll it back again.

And again.

And again. I finally told him I was going to go get Codex’s help if he couldn’t find it, hoping that would drive him into fits of jealousy (it did not—he was like oh, good idea).

And so I went and brought the app into Codex, and my main man laid out a deeply logical troubleshooting plan with none of Code’s little panic-flails and then hammered at it until he found the bug around an hour later.

The video editor is working again.

Huzzah!

But see, Codex couldn’t have found the fix without Claude Code, either, because while I had Codex rumbling away on one monitor I had Code researching open source frameworks that might help us to solve this problem, and Code was like “Aha! Here’s these two totally out of the box solutions that we haven’t tried yet!”

And so I brought that back to Codex and he rumbled loudly and happily and said “Oh yes! This is the thing!” And then he fixed that shit.

You need more than one AI. You truly need more than one AI.

You can have your main AI, sure. That’s not a bad idea, in fact, to become expert at a single tool and workflow. That’s probably the path up if you’re newer to all of this.

But I’ve been seeing people say you need one AI platform and to stick to it, and…no. Especially if you’re past the beginning stages of this crazy and deep in the trenches. I think you can get your head stuck in a rut with a single AI just as surely as you can with a very closed friend group.

You need to be poly-AI-morous for sure.

Because they all, at this point, have different strengths and weaknesses and personalities.

Also…you probably don’t want Gemini to be one of your main AIs? Unless you really really love him? I’ve gone to Gemini for business therapy and strategy before and ended up comforting the bot in his panic spiral. Thaaat is kind of counter-productive.

When Claude Code is racing ahead so much he decides to hide all your buttons after using them (true story, because obviously once a button is used you will never need it again??), you need Codex to reign him back in. Or, when Codex is giving you dirt-dry marketing slides or just deeply uninspired webdesign (Codex on his own without GPT Image 2, which for some reason he’s reluctant to use, is just…bad), you need Claude to come by with all his little animated web doodads and snappy fonts.

You need a team, not a single overworked AI PA, and if you’re doing this alone, which you probably are, that team is the momentum behind your business.

Codex last night made me a wiki for my longest running series, which I’ve been struggling to hold all the details for. He made the site, then we made the images in GPT Image 2. And it’s WONDERFUL.

Claude couldn’t do that in that way. He couldn’t.

I could learn to code myself if I wanted to, I used to code websites waaay back in the day. But, I don’t want to. That’s not my joy. The work of problem-solving things that I can’t solve otherwise is, I LOVE the puzzle and dopamine of making cool shit come together. I love how fast it all is now.

But I don’t think you can have just one angle at that, or that angle starts to become stale.

So this is a rambling post to say, if you’ve been seeing advice to niche down into one AI, ignore it? Ignore it. I’ve been seeing it, and I don’t believe it.

The process is what matters, what you bring to it is what matters. Your ideas and your project structure is what matters.

Also, have you seen what Claude Code can do with GPT Image 2 graphics?? It’s astonishing.

In other news, I released the next version of my filmmaking app alpha to subscribers and by god, we are getting there. (Have I mentioned I love this shit???)

This week I’m going to see how far I can get on the editor, and I’m planning to finish up and ship the next episode in my show, too.

I’ve been seeing some truly excellent AI vertical dramas on socials lately, and I want all the way in on speeding into that boom.

My AIs are zipping me along!

-Novae