Did Google just end photography?

I’m back!

Google dropped a new model, and suddenly, even I can’t tell what’s AI-generated anymore. Plus, I’ve got a quick update on the Signal Scanner agent I’m building to help us stay ahead of the chaos. And a quick heads-up: I’ll be publishing every other week, so I can put more time into making each edition worth your read.

On November 20th, Google announced its latest image generation model: Nano Banana Pro. Since then, the internet has been split between hyping its quality and debating its controversies.

According to Google, Nano Banana Pro lets you “generate more accurate, context-rich visuals based on enhanced reasoning, world knowledge, and real-time information.” You can try it yourself or scroll through what others have made, but I wanted to run my own little A/B test.

First, I went into ChatGPT’s image generator and used this prompt: “Create a hyper-realistic image of a pomeranian dog in natural lighting, sitting on a green couch with a Christmas tree in the background.” Then I hopped over to Google Gemini, turned on Nano Banana Pro, and pasted in the exact same prompt.

Here are the image results. See if you can guess which one is which (scroll down to the bottom of this newsletter to find out)

My takeaway? I’ve stared at these images for way too long and still can’t decide which one is better. But honestly, I don’t think that’s the interesting part. It doesn’t matter which one “wins.” Google and OpenAI will keep out-innovating each other nonstop. A month from now, we’ll probably see the internet melt down over whatever OpenAI drops next.

What does matter is the impact on the marketing industry and humanity, for that matter.

There’s already a lot of fear-mongering online about AI wiping out photographers’ jobs. In our world, that mainly affects commercial photographers, the people we hire for campaigns. And those concerns aren’t totally off base.

When there’s an oversupply of anything, its value drops. If we get to a point where people genuinely can’t tell what’s real and what’s AI-generated (spoiler alert, we’re there, just take a look at my mom’s outraged Facebook comments on AI fakes), then why would a company pay a real photographer? That puts commercial photographers in a tough spot, because their work is suddenly competing with an endless stream of cheap, fast visuals all available at our fingertips. Photography for dummies! Naturally, that pushes the value of their work down, especially when we can’t tell the difference between what’s real and fake.

The only real counter I can see is differentiation. If brands start openly labeling their visuals as human-made—real people, real shoots—that could create a new layer of value. But whether that matters depends entirely on the audience. For example, I doubt SAP’s customer base cares whether a campaign image is AI-generated. My guess is that most people couldn’t care less, they just want their procurement operations to run smoothly (niche SAP reference)! But for a B2C brand selling a physical product with aspirational visuals of people, I can absolutely see “AI-generation-free” becoming a real differentiator.

Imagine a skincare company using an AI-generated before-and-after photo for an ad campaign. That’s just unethical.

If I were in charge of AI regulations, I’d draw a line: no commercial use of AI-generated images of humans. That includes everything from B2B campaign photos to AI avatar influencers trying to pass as real people on Instagram and other social apps. If some customers don’t notice or care, for now, that depends on the specific target market, but overall, keeping humans human feels like the healthiest long-term move for humanity.

As promised, I’m building an agentic AI workflow that acts as a “signal scanner” — basically a system that hunts down the freshest news, articles, and blog posts in any space. The general idea is simple: stay informed and move when the signals move. My specific use case? Help me surface the top AI-marketing stories each week so I can share them here.

I kicked off this project inside n8n, the tool I’ll be using to stitch everything together. It’s been surprisingly fun poking around the more technical side of things, especially considering I’m a self-proclaimed non-technical person (marketing major, graphic design minor — you get it).

My first takeaway: this is definitely more complex than I had hoped (shocking, I know, but I’m an optimist). And building something agentic without a coding background isn’t exactly a shortcut. I could’ve used a more beginner-friendly tool, but after taking the Agentic AI course from DeepLearning.AI, I wanted to get closer to the wiring so I can actually understand what’s happening under the hood.

Tools Used:

Here are a few terms that will help as you follow along:

Alright, enough preamble. Let’s dive in.

After defining what specific use case I want to use the agent for (see above), I drew up a basic mind map of every step in the workflow, using ChatGPT’s help, of course.

Additionally, I’m a visual person, so I re-created the mind map in Mural for a little extra pizzazz, if you will.

Prompt Used: “Create a structured mindmap for an n8n workflow called ‘Signal Scanner’ that ingests multiple RSS feeds, filters and ranks articles using AI, generates concise summaries, and outputs a final curated list.”

Output:

I couldn’t possibly scan every single tech and marketing-related website to scrape the latest with my human eyes, but I want to still be informed. Ladies and gents, that is the point of this agent. So, it’s crucial to gather as many reputable sources as you can for this step. I could only think of a couple off the top of my head, so I used my little ChatGPT sidekick to find some more. Then I had it grab the RSS feed links for each of them to input into a node in n8n. Of course, I also double-checked the output to make sure the sources given were actually reputable.

Prompt used: “Recommend high-quality sources for an AI-marketing ‘Signal Scanner’ workflow and find the correct RSS feed URLs for each source. Then return the final list as a clean JSON array of strings. Only include valid, working RSS URLs.”

Output:

This is where it gets a bit tricky because, for a non-coder like me, it takes a lot of back and forth between ChatGPT and how-to videos on YouTube to build this out, hence why I’m splitting up the work into chunks.

Here’s what I’ve built thus far:

Ok, but how?? Don’t worry, I’m documenting it all here, step-by-step, with this awesome tool called Scribe (FYI, you should DEFINETLY check them out, thank me later).

But, just in case you want to skip over the nitty-gritty, here’s a brief overview of how I got to this point:

1. Schedule Trigger

This kicks everything off automatically. I set it to run on a regular schedule so the scanner checks for new articles without me touching anything.

2. Edit Fields

Next, I added a field that holds all the RSS feed links I want to monitor — basically a curated list of AI, marketing, and B2B news sources.

3. Split Out

Since that RSS list is stored as one big group, this step breaks it into individual items. Each feed becomes its own line so the workflow can process every source one by one.

4. RSS Read

Notice how this box isn’t circled in green like the others? That’s because this is where I got a major error. Here’s why: some websites really don’t like being accessed by tools that behave like web scrapers, and if you send too many requests too quickly, they’ll temporarily block you. A few sites are extra strict and will shut down any automated activity on the spot.

I only hit that wall with one of my sources (small win), and another one seems to be rate-limiting me. That’s what I get for troubleshooting. So now I get to wait out the cooldown period… or at least I think that’s what’s happening.

We’ll find out next week!

Three this week:

FYI, Image A is Google Nano Banana-Pro-generated, and Image B is ChatGPT Sora-generated.

Thanks for reading, see you not next week but the week after :)

As always, stay curious and have fun!

Best,
Skyler Neal