Love this newsletter? Forward it to one curious friend. They can join in one click.

Welcome back to a SPECIAL EDITION AI For Humans newsletter!

Google’s new AI image model Nano Banana Pro is here as of this morning.

And it’s VERY GOOD.

How good? Well, here was my first output:

If you haven’t been following AI image models that closely, you might look at this and be like ‘huh, cool but not that exciting’.

But as someone who’s spent years toiling in the AI image model mines, I can tell you the above prompt is insanely hard for most of them to figure out.

It’s the scope of the ask, the differentiation amongst the signs and the consistency across the whole thing.

That’s where Nano Banana Pro shines. With Gemini Pro 3’s reasoning model baked in, it can handle significantly more detailed prompts, image edits and fails a whole lot less.

I’ve been playing with it all morning (and will share some of my favorite prompts from others below) but here are some basic things you need to know about the improvements:

Maybe more than anything, the model just feels like it can take on more stuff and give you better outputs. It mostly just works.

All that said, it’s not perfect.

If you’ve been in our Discord, you know that I’m a big fan of the Dungeon Crawler Carl series and I asked it to make me an infographic of the first six levels of the Dungeon (iykyk).

You can see that while there is a TON of good info here, some of the words are still smoosh-y and illegible. But that’s a LOT of words.

We’ll have a bit more on this in tomorrow’s AI For Humans (spoiler: we recorded last night because of a travel issue) but again, you can do use this right now. It’s baked into Gemini or you can try it at the AI Studio as well.

Tomorrow, we actually spend a lot of our time focusing in more on the other Google craziness that happened this week in AI: Gemini Pro 3.

I won’t go into much detail on Gemini 3 Pro here (watch the show tomorrow) but spoiler: It’s GREAT.

The basic tl;dr is that Gemini 3 Pro is far and away the current state-of-the-art AI model and kind of shocked me at just how much better it is than most everything else out there.

For our nerdier readers, below is the best Youtube video you can watch about all the technical upsides of Gemini 3 from our favorite (non-AI For Humans) YouTuber AI Explained.

Of course, OpenAI wasn’t going to take this launch sitting down. Yesterday, they unleashed an updated GPT-5.1 Pro and GPT-5.1 Codex MAX (their coding platform) which, it turns out, is ALSO very good.

If you’ve been wondering if we’ve hit a wall in the AI space (or people you know have been saying that), I think the last week is a good example of just how fast stuff is changing and improving.

See you tomorrow for the show!

- Gavin

One of the best things when a new AI image model launches is seeing what everyone else is doing with it.

Below are some of our favorite examples so far:

And, of course, Mr. Bean in Total Recall: