In this week’s newsletter, I bring you on a quick tour of Google’s new Gemini 3 for vibe-coding and the phenomenal Nano Banana Pro imaging engine. This key visual was generated by NBP within Google AI Studio, with an interesting combo of traditional painting techniques and 3D lighting.
When it comes to using Gen AI, I’ve observed that many people only know and use ChatGPT. How much they are missing out on, even as free users! For the past year, Google’s Gen AI feature set has far surpassed OpenAI’s, and it is time for you to explore them.
In my AI workshops, I spend the first few hours teaching advanced ChatGPT skills, and then the majority of the day is spent within Google’s Gen AI ecosystem – Gemini, NotebookLM, Whisk, AI in Google Sheets, and increasingly, Google AI Studio.
This past week, Google trotted out the latest Gemini 3 model and Nano Banana Pro, and both are a leap forward in LLM reasoning, vibe-coding and image generation. I’ll talk less and show you more. Warning: Many visuals ahead!
Gemini 3 vibe-coded apps
It’s complicated to prove how Gemini 3 is better at reasoning for daily prompting, so I’ll show you the vibe-coded apps instead. You can build simple apps in Gemini, and more complicated ones like these two in AI Studio:
My RhythmKids AI app
I generated this rhythm beatbox machine for kids. I originally designed a drum machine but found it too complicated, so I made this instead. You can click the boxes to add drum beats and snares, or just generate with pre-made settings. FYI, it took several tries before the app would work properly, so be patient with AI when vibe-coding.
The “Next Token Predictor”
I also improved the interactive demo which I use to explain how Gen AI generates every word using probability and not actual “intelligence”.
You’ll read a lot of articles about how vibe-coding is not useful or dangerous. Both statements are true for many use cases, but if you are careful about data security and have real problems to solve, vibe-coding is going to change your workflow in yet unimagined ways.
Nano Banana Pro
I started my Gen AI journey in 2022 with the janky Stable Diffusion image generator. If I had hair to pull then, I’d be balder than I am now, because it was impossible to wrangle viable images out of SD without a lot of plug-ins, GPU horsepower and patience. SD is a faded memory now, as ChatGPT and Gemini leapt ahead in generating visuals.
In 2025, Nano Banana Pro is breathtaking on many levels. You can use it within the Gemini app for free (limited generations), but if you are willing to pay per image, Google AI Studio lets you generate up to 4K resolution and select the aspect ratio. It’s quite affordable – each 2K resolution image costs about USD 13 cents, and each 4K image is 24 cents.
Here’s one of the draft designs for today’s “Iterate This!” newsletter banner.
See content credentials
Very Arthur C. Clarke.
Then I wondered what would happen if I threw my CV into Nano Banana Pro…
I like the icons here!
Hmm, something less boring?
At this point, I realised I had to upload my bald photo to depict myself.
What if I made my career into a children’s storybook?
See content credentials
OG means that I’m the original Ian Tan. Get away from me, the other Ian Tans on LinkedIn!
Ok, let’s make it more LinkedIn-ey:
See content credentials
I made this as complicated as I could to test how far I could push the Banana engine.
Reimagining my ThinkTan company logo
Plain and simple, or just lazy design? That’s my logo!
My company logo is terribly plain, largely because I only took 5 minutes to design it so I could have a visual that was scalable for different uses.
Now look what happens when I transform the Banana way:
This took quite a bit of iterative prompting to get the surface materials right.
See content credentials
Inspired by “The Expanse’ TV series.
See content credentials
Corporate branding on a stellar scale
At this point, some people will say that graphic designers have been wiped out by Gen AI. That’s only true if you are a weak designer without a strong design foundation. Strong designers will find AI to be a great partner in ideation and prototyping.
For me to create each of these visuals, I had to bring in all my design skills – image ideation, mixing concepts, mental visualisation and directing the AI to do what I wanted. With or without AI, a designer needs years of experience and domain knowledge to turn thoughts into visuals accurately and vividly. I could have also uploaded my sketches for AI to transform them, but I’ll try that another day.
Anyway, I’ve been busy this past week with talks and workshops (Seng Kang Secondary School, Nanyang Business School) , so I haven’t had time to explore Gemini and Nano Banana Pro further. So, I’m only scratching the surface and you should start doing so!
Other stuff I wrote this week
I finally assembled my Frankenstein PC, a geek project I’ve been thinking about for 10 years.
Coincidentally, Valve announced the Steam Machine, which is a tiny PC which will disrupt the PlayStation business big time from 2026. I’m not mentioning Xbox as it has utterly lagged in the home console wars. I’ve believed for decades that the PC will defeat all consoles, and here’s someone else who thinks so.
I have the writer’s affliction of noticing poorly-written small print and thinking incessantly about it.
ChatGPT helped me to identify the world’s dumbest kamikaze bug.
That’s all folks, thanks for reading!
23-Nov-2025