Superpowers from Gemini – Part 2
After last week’s post on how to get real superpowers from Gemini, I received several messages and questions.
Thank you all.
Some made me reflect. So I’ve decided to write this follow-up to answer a few of the most common ones.
(And if you missed last Friday’s post, I recommend reading it first—it’ll make this one much more useful.)
Let’s go for it.
What about context?
A key part of the approach we described is creating a single, pinned chat, assigning it a clear expert role, and feeding it with information: documents, transcripts, notes, etc.
But many of you asked: Won’t it forget? What happens when the context window fills up?
And the answer is: yes, it can forget.
One million tokens is a lot—roughly 20 books—but if you’re working on a big project, you’ll eventually lose some of the early context.
So here are some trick you can use:
Start with a Gem.
When defining the role, instead of just copying the persona into a chat, you can first create a Gem with that role and its full description.
Then, open your main chat based on that Gem and make that your pinned expert chat.
This way, the role and personality are permanently remembered—even if the chat eventually forgets some of the other content, it won’t forget who it is.
You can also use this Gem for other tasks that don't require context.
Why not use Notebook LM?
Another recurring question.
Short answer: I do use it. A lot.
Notebook LM is excellent—but you have to understand what it does and what it doesn’t.
It doesn’t create new content. It’s not a creative assistant.
It’s more like a super-smart Q&A system that helps you extract, explore, and summarize what’s already in your documents.
So here’s how I use it in combination with Gemini:
If I have a big chunk of content—like long PDFs or background docs—
I upload them to Notebook LM and ask it to create a comprehensive guide or summary.
Then I copy that summary into my Gemini expert chat.
This gives Gemini a clean, compressed version of that content without overwhelming the context window.
Also useful for extracting structured info from websites—just paste the URL into Notebook LM, grab the insights, and paste them into Gemini.
What about learning about tools or vendors? Try Deep Research.
If you’ve never used it, you’re missing out. It’s a research method that runs for about 15–20 minutes and returns a detailed 10–15 page report on any topic, including feedback, analysis, pricing, and more.
Don’t run the Deep Research in your main expert chat.
Do it in a separate one. Research any topic you want to learn about and that can be useful for your project.
Then copy and paste the best parts into your pinned Gemini expert.
It will then “digest” that knowledge as part of the project’s brain.
Can it create actual documents?
Another great question.
The short answer is: yes—sort of.
If you ask Gemini, “Write the communication plan for this program using this template structure,” it will generate a very solid draft with sections, key points, and consistent tone.
You can then copy and paste it into Google Docs, tweak a few lines, and have a version 1.
You can then discuss it with your team, get feedback, record the conversation, paste the transcript into the chat and ask to generate version 2.
You will see real magic doing this.
In many cases, the final format matters less than having the right structure, content and feedback incorporated. Templates are useful—but they shouldn’t be obstacles.
What about automations?
There’s growing curiosity about task generation, automated updates, and connecting this flow to real tools.
I’m exploring that now and will probably write a future post on it.
For now, I just want to say thank you again for all the comments.
If you have more questions, ideas, or use cases, just write me a message.
I’d love to keep the conversation going—and unlock even more superpowers together.
“The mind is not a vessel to be filled, but a fire to be kindled.”
— Plutarch