Get comfortable switching models

The best AI workflows are built on adaptation, not allegiance.

If you care about results, and you should, there’s no absolutely no advantage to staying loyal to a single AI provider or model.

For the past few years, image generation has mostly been… fine. Good enough for announcement reels and flashy website demos, but rarely quite right when you needed something truly usable.

Earlier this year, Nano Banana 2 was released and earned its reputation as the image-generation king.

At the end of last month, OpenAI released gpt-image-2.It took me a while to even notice it was there because I’d defaulted to Gemini + Nano Banana for images.

But after using it extensively over the past week (especially for Substack thumbnails), I’m honestly blown away by the quality.

They’re so good they make me want to create YouTube videos… just so I can generate thumbnails for them 😅

Better yet, it’s integrated into Codex (OpenAI coding agent), so you can generate assets while you’re actively making changes to an app or website.

Just today, I had it update a hero background image based on new landing page copy, and it captured exactly what the client wanted to convey.

Marketing copy → Image. No separate prompting workflow. First try. No back and forth.

Talk about productivity!

And sure, something even better will probably come out in a few weeks. Great! Bring it on!

That just means I’ll have an even better tool to help deliver results.
That’s all I care about, and you should too.