AI Field Notes: Google is Crushing It
I switched back to Gemini, and you might want to too
Every time AI improves, more use cases become viable. Our AI Field Notes posts track the practical, near-term applications emerging from our day-to-day work. While some follow Wayne Gretzky’s advice to “skate to where the puck is going,” we believe it should be balanced with rapidly skating to where the puck already is. We build flexibly to capitalize on current and future opportunities. This series highlights what you can implement today.
Google’s last two weeks have been extraordinary. Google is crushing it. This is where we are in the two year cycle:
Nano Banana Pro
Google’s “Nano Banana” was already considered best-in-class at image generation, but the new model surpasses its predecessor. Where’s the unlock?
Fidelity to Text: It captures text nearly perfectly in standard scenarios.
Visual References: You can incorporate up to 14 visual references into the image with high fidelity.
Image Quality: It generates professional-level 4k images.
Logical Consistency: It has the ability to follow complex instructions; even complicated diagrams can be constructed accurately and consistently.
What are all of those use cases for AI generated images you’ve thought of?
Custom flashy slides
Detailed marketing posters
Illustrations with consistently portrayed characters in a story or comic book
Any job that usually takes 30 minutes or more in Adobe Photoshop
Any touch up in Photoshop that takes less than 30 minutes, but doesn’t have to be exactly perfect.
It can do it all.
I turned most of my recent blog post for Pattern into a magazine shot, word-for-word:
Google Begins Swapping to AI Default for Search
Consider the following:
Google has more users than any of the other large tech companies.
Google makes insane amounts of money from search, basically acting as a monopoly on advertising search content.
It is not obvious that 10 years from now, Google Search in its present form will be a viable business.
There is no business model YET within AI that is as profitable as the current Google Search business model.
Google’s “self-disruption” decision to swap the default mode from Search to AI is one for the ages. They are potentially lighting hundreds of billions of dollars on fire to bet on this transition. Will they do it?
Google just started rolling out default AI mode to some users.
Old default view for Google.com:
New default view for some desktop users on Google.com:
Gemini 3 – Use It Now
Chat GPT 5.1 was released to some fanfare. It looked great in benchmarks, but the quality immediately dropped for all of my [non-coding] use cases. The reasoning would go haywire on me on a relatively frequent basis, including this fun instance where it tried to correct the spelling of “shoos.”
Driven by my frustration with 5.1 and the buzz surrounding Gemini, I came back to Google’s model, and it quickly became my go-to for most use cases. Yes, in a month or two someone else will come out with a breakthrough model and we will all switch again. But if you have access to Gemini Pro, you should start leveraging it today.
Shareable Gems Going Mainstream
Now that Gemini Gems are shareable, one of the easiest corporate AI enablement hacks is to quickly stand up a knowledge management system for a team, department, or company.
Here are the steps:
Go to Gemini
Create a Gemini Gem
Drop in the company documents, guidelines, policies, and best practices you want accessible to the team.
Share the gem with the team, just as you would a Google Drive folder
For a group of 5 to 150 people, this is the easiest way to stand up a knowledge management system. Larger groups will require more controls and scaling methods to build high-quality infrastructure, but this is the best solution for any small group right now.
Our Parting Thought
Find what’s useful. Discover lasting principles. Be wrong. Adapt.
