Hello Full Stack PMs!

Welcome to the fourth Weekly Stack, where we concentrate the AI news firehose into tactical insights for PM builders. We've got 270 new subscribers this week – welcome to the stack! 🥞

After getting Gemini 2.5 Flash and realtime-gpt last week, and GPT-5 shortly before, we’re finally getting a breather. There really wasn’t a lot of news, and you can trust me not to hype anything up just to hype it up.

I still have some great things to share with you:

  • A powerful tool I launched for subscribers called Product Pulse 💛

  • Follow-up from last week with examples and tons of resources for ways to use Gemini 2.5 Flash / Nano Banana (once again featuring my cat Piper)

  • An emerging trend that will hit hard in the next month: AI Everywhere (it’s really happening)

Let's do this.

💛 Resource I Made – Product Pulse

There’s a lot of product content and AI news, and keeping up with it can feel like a full-time job. (It practically is my full-time job.) So like a good Full Stack PM, I tried to build something to help me get a handle on it.

I came up with Product Pulse 💛

Here's what it does (demo video):

📰 Scans 20+ top-tier product & AI newsletters
⚡ Adds bullet-point summaries to a database every day
🔍 Lets you filter by date, tag, free/paid
📬 Can email digests directly to you

It’s just for Full Stack PMs. 100% free, of course. Here's the link.

Next week, I’ll share how it’s built and open-source everything so you can build your own. I’ll show you how to add your own sources, modify the summarization prompt, and even hook in a Slack bot to DM you highly relevant news.

It’s less scary than it looks.

If there are any great newsletters or blogs you think should be included, let me know!

🍌 Follow-Up: Tips & Tricks for Nano Banana

As I covered last week, the mysterious, incredible image gen model code-named Nano Banana turned out to be Google’s Gemini 2.5 Flash. But the internet is so in love with the name Nano Banana that even Google is leaning into it.

I’ve been playing around with it a lot, and these are the best tips and resources I’ve found.

First – product use cases for Nano Banana

Yes this technology is amazing, but why is it relevant to PM builders? Nano Banana is perfect for character consistency, realistic edits, and image-to-image fusion.

That makes it great for product storytelling across multiple scenes and quick iteration.

Some examples:

  • Storyboard user journeys: Use the same persona across panels to illustrate user journeys.

  • Before/after concept comps: Turn sketches/whiteboards into polished “future state” visuals.

  • A/B GTM creative: Spin on-brand variants while holding character/device constant.

  • Prototype realism: Drop early UI into lifelike scenes (in-hand phone, desk) for user tests.

And this doesn’t even begin to cover all the non-PM use cases people are coming up with. It’s too much to include here, but just take a quick gander.

Use JSON inputs + reference images

JSON is is just a structured style way of organizing information. e.g.:

{
  "cat1": {
    "name": "Winter",
    "fluffiness": 6,
    "crimes": ["knocked plant off table", "ate my lunch", "3am zoomies"],
    "currentMood": "plotting",
    "treatsEatenToday": 47
  },
  "cat2": {
    "name": "Piper",
    "fluffiness": 9,
    "crimes": ["screamed at wall", "hid in Amazon box", "tore up the trash"],
    "currentMood": "baby",
    "treatsEatenToday": 3
  }
}

LLMs love this stricture. It’s great for image generation since it gives you fine-grained control over the variables.

Here are the variables you can control, with guidance for each:

And here’s an database I found with 100+ examples in this format. They are more marketing oriented, but show the idea. (We could use a better database for PM use cases specifically… hmm I’ll write that down.)

Just choose one and plug it into Google’s AI Studio.

It even made those modern wordmarks.

Actually building with Nano Banana

If you want to try integrating Nano Banana into a project, my friend Ricky Kirkendall built this great open-source tool called nano-agent that lets you create images directly from any coding environment like Cursor, Windsurf, or Claude Code.

Just paste the GitHub link into your AI of choice and ask it what to do.

📈 Trend to Watch This Month: AI Sinks Its Teeth Deep

We’ve been hyping LLM integration ever since the world discovered ChatGPT 3 in 2022. September 2025 is when we’ll see it start happening for every device you own.

Apple's jumping in next week (September 9th) with their "Awe Dropping" event where they'll finally show off the iPhone 17's AI features – maybe even a Gemini-integrated Siri that actually... works?

Meanwhile, Google's Pixel 10 is already shipping on-device AI that does stuff like proactive assistance with data from all your apps (Magic Cue) and real-time camera coaching without even touching the cloud. Phones are getting their own AI brains.

AI isn't just a feature anymore – it's becoming the foundation layer of everything. OpenAI's new Realtime API means any app can have natural voice conversations. Anthropic just hit 300k business customers and went from $1B to $5B in revenue in 8 months (!).

What this means for PM builders:

  • Users will expect AI in everything in 2026. Not having it will be like shipping without search in 2010

  • The integration layer matters more than the model – pay attention to good AI UX wherever you see it

  • Start experimenting now with voice interfaces, multimodal inputs, and proactive assistance

Of course, I’ll be exploring all of this with you to help you stay ahead.

😄 Meme of the week

Related to the trend above:

📚 Other good reads & listens

  • Tools Ben of Ben’s Bites actually uses daily: Great article from Ben’s Bites where he gives an overview of the AI tools he uses daily. Super tactical stuff.

  • Vibe Coding 101: 23 practical tips to build functional prototypes: This an absolute monster of an article. It’s huge but fast paced, jam packed with great tips from beginner to very advanced. If you follow this whole article, you’ll be at the pinnacle of vibe coding.

  • OpenAI starts opening up their roadmap: OpenAI is rolling out automatic detection and routing of mental health conversations to specialized "reasoning" models that think, plus Parental Controls for teen accounts within the next month. This is part of a major shift of AI companies proactively building safety rails and age-appropriate features before regulators force them to. This could set a new baseline that all consumer AI products need to match.

🥞 The Last Pancake

If you only have 30 minutes this week, try out one of the prompts from the JSON database above and make an image in AI Studio!

Keep building,

Carl

How did you like today's newsletter?

Login or Subscribe to participate

Winter says her only crime is being too cool.

Keep Reading

No posts found