The Masked Medici: How to Build a Faceless Youtube Channel and Companion 1990s Strategy Game in a Single Afternoon with Google AI

Today in this special operator's bonus episode, we are going from zero to a full multimodal app experience set in the Renaissance with Gemini, NotebookLM, and Google AI Studio. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, we have something a little bit different and quite a bit fun today. A conversation recently came up with Google where they were interested in exploring some sort of collaborative sponsored partnership style episode. Now, one of the things that is of course most unique about Google is just the sheer breadth of products they have in the AI space. In fact, there's so much that sometimes folks don't even realize how much is actually available to them. And believe it or not, as the conversation had started, I thought back to an idea for a project that was one of those ideas where, you know, there's so little reason to do it and so many other things that need to be prioritized in front of it. That you really shouldn't be spending time even thinking about it, much less actually considering doing it, and yet it gets in your head like a little brain worm that just won't go away? Well, one fact that you might not know about me is that I'm an absolute history nut. I was a history major. I never really considered majoring in anything else. And at any given time, I always have some history book that I'm reading. And for the past 4 or 5 years, every time spring starts to turn into summer, I always find myself gravitating back to the Renaissance. This maybe will be less surprising, but I am completely fascinated with liminal moments, these moments in between big epochs of history. And the Renaissance was of course one of the most profound of those types of liminal moments that we've ever experienced. It was the bridge between the medieval and the modern period with much of what would lay the foundations for the next 500 years of history started in just a few short generations. As I've watched others experiment with AI, the one thing that I kept wanting to do just for the sheer joy of it, was to create a faceless YouTube history channel focused on telling some of what I think are the most interesting stories from the Renaissance. Now, I actually had already thought a little bit about how you would wire together a bunch of different AI services to actually automate big chunks of this and make it viable. And it turns out that at this point, Google and Gemini have pretty much all of those in-house. So as we started talking about this idea for this episode, I pitched them on this, and here we are. So what we're going to do today is walk through the sequence that I used to ultimately produce that faceless YouTube channel, which I'm calling The Masked Medici, the companion website for their YouTube channel, and even an early 1990s style strategy game based on the same themes. So let's walk through how we did it. First up, of course, was setting up Gemini to be our build and creative partner for this entire experience. First thing I did is set up a Gem to keep all those conversations in one place. And then we started to brainstorm. Now, like I said, I had this concept for this Renaissance-focused faceless channel right from the beginning. And so where we started instead was thinking about all the different things we could put around it, plus all the different tools we could pull in. One of the things that I wanted to brainstorm first was how to get the most value out of some of the most unique parts of Google AI Studio, particularly its ability to integrate Gemini's AI and multimodal capabilities into whatever it was that we built. No boring text-based websites for us. Pretty soon we started to hone in on some sort of interactive experience as a companion to the YouTube. And as we started to refine what the content for both the companion web experiences plus the YouTube channel could be, I also switched Gemini out of its architecture mode into its historian mode to think about a few key historical moments to focus on for this channel's launch. So to keep track, the first Gemini capability that we're using is effectively all the dimensions of its strategic planning with a little side of its historical knowledge, although very quickly we turned and doubled down on that specific knowledge and the capability to research church and go do more. I knew that one of the stories that I wanted to focus on was the Pazzi conspiracy. This was this incredibly dramatic moment when the feud between the Pazzi and Medici banking families came to a head with the attempted assassination of Lorenzo de' Medici, who would eventually go on to be known by history as Lorenzo the Magnificent. Although Florence technically had no ruler, it being led instead by a very proto-democratic body called the Signoria, there was no doubt in anyone's mind even by that time in 1478 that the Medicis were the true rulers of Florence, the power that ran the city even if from behind. Now, part of what made the Pazzi conspiracy more than just a bloody feud between families is that the plot had the backing of the papacy under Pope Sixtus IV. And as if that weren't enough, it also had the military support of the Kingdom of Naples. And yet, while the day extracted a heavy toll, with Lorenzo's younger brother Giuliani being killed in the fray, Lorenzo himself survived and would go on to consolidate power in a way that the Pazzi family would never recover from. Now for this, we actually did deep research in two different places. I used the deep research feature that is directly in Gemini, but where the real power came from was NotebookLM. With NotebookLM, you can go find and integrate dozens and dozens of sources about any topic that you're interested in. You can give it your own sources, which can be anything from uploaded files to websites to YouTube videos to access to a Google Drive folder, or you can have it help you go get new resources from the web. With either a fast research or a deep research mode. For each of the 5 notebooks I created for what would become the 5 videos, I used the deep research feature to go get the set of sources that we would pull from to actually craft the final content. Now you might remember that NotebookLM first started to break out at the end of 2024 and beginning of 2025 because of its audio overview feature. You could take a bunch of sources or even just one source like a big dense AI research paper and with a single click turn it into a podcast where 2 hosts talked about it with all of the affectations of your favorite conversational podcast. And while this is the type of feature that could become a novelty fast, part of the reason that I think it was so sticky is that it's actually just kind of a good way to absorb information. And so for people who were trying to learn about new topics, it became a useful tool in the arsenal. Now since then, however, Notebook LM has continued to push new features and capabilities. In particular, the increased capability set around visual generation has made a huge difference in what you can do with Notebook LM. As we started to get both the text rendering capabilities as well as the reason over image capabilities of the Nano Banana models, that opened up new features like infographics as well as slide decks. And one of the things that many people found, especially with the infographics, is that because you can curate such a bigger set of sources, the infographics that get produced by NotebookLM are often much better and more factually dense than even the versions produced in the regular Gemini app. Also taking advantage of those capabilities are the slide deck builder in NotebookLM, which as of February, you can edit on a slide-by-slide basis, which some folks like Click Health's Simon Smith called a death blow for many AI presentation generators. Still, the feature that I was focused on for this particular project, and frankly, the thing that made my longstanding idea of a faceless YouTube channel actually at least a little bit viable, even despite everything else I have going on, was that at the beginning of March, NotebookLM added what they called cinematic video overviews. As they describe it in the app, these are rich immersive experiences that can unpack the complex ideas of your sources through engaging visuals and storytelling. It was those cinematic videos that would be the substance of our YouTube channel. And so we built notebooks for our 5 topics. There was the Pazzi conspiracy, Lorenzo's dramatic flight to confab with the King of Naples, Brunelleschi's dome, a marvel of architecture, science, and art that pretty much no one at his time could understand. The Bonfire of the Vanities, which signaled the end of one period of the Renaissance and the beginning of something very different. And of course, the attempted mutiny against Cesare Borgia that would go on to inspire George Martin's Red Wedding. I won't play the whole video, but I want to give you a sense of what these cinematic overviews are like.

Florence, December 6th, 1479. It's the dead of night, and the unofficial ruler of the Florentine Republic is sneaking out of his own city. Before leaving, Lorenzo de' Medici left a letter for the Signoria, the city's governing council. In it, he wrote that the enemy armies currently ravaging the Tuscan countryside were driven by a singular hatred directed entirely at him. His plan was to sail south, directly into the capital of the enemy coalition. He was going to surrender himself to the King of Naples to either negotiate a peace or face execution. With Florence on the brink of total collapse, Lorenzo calculated that offering himself as a willing hostage was the only remaining play to save the state.

So a couple things that I think are worth noticing here. First of all, the cinematic video overview is actually using a couple different types of image sourcing. Like, for real-world architecture, like this photo of Florence, it's pulling from and referencing specific licensed stock photography. Yet of course, for a lot of this, they're using a combination of Nano Banana 2 and their VEO video models to create the images, both still and moving, that make up the substance of most of the video. What particularly impressed me about this is that it chose a visual style and stuck with it for the whole video. All of these images, this sort of thick, laden oil painting style with big visual brushstrokes, look like they go together. It doesn't look, in other words, like some random assemblage of stock photos. It has a consistent visual identity. So okay, we now had our raw material. Now of course, where this was going to land was on a new YouTube channel, of course also a Google property, and everything about this from the name to the images to the video names, the descriptions, and the companion text, all created by Gemini or Nanobanana. But what about the web companion experiences and the interactive elements? Well, in the conversation with Gemini, we'd come across two ideas that I thought were worth pursuing. The first was some sort of digital codex illuminated manuscript style presentation that would house the actual videos themselves. You can see Gemini writes, this is a beautiful, elegant, and highly achievable entry point. It frames your content not just as YouTube videos, but as pages in a grander historical text. That was idea one. Idea two, which would evolve a little bit, was called at the time Florence and Factions. A 1990s-style computer strategy game that would actually bring you back into that period and have you act as someone living in Medici-era Florence. Now, this is where we would use our next tool, which is the recently updated Stitch. Now, this update at the time of recording is exactly a week old, but already I had seen so many people doing cool things with it that I knew that I wanted to try it for this particular project. Stitch is nominally a design platform, but it's got this endless canvas and creates not just images but entire design systems. So for example, the first of these we did was the web app to house the videos. It came up with this visual motif, the name, the use of the written out Latin year, as well as the whole color and font system. Now as you'll see in a minute, part of what makes Stitch really powerful is the way that you can go back and forth and iterate in one canvas, but this kind of one-shotted pretty much exactly what I was looking for. And once you've got something you can like, you can export it, you can bring it into Figma, You can turn it into an MCP, you can create an instant prototype, or, and this is what I did, you could bring it into AI Studio. So after I clicked export, I selected AI Studio, build with AI Studio, and it immediately brought me into the Google AI Studio experience where it automatically came with the image, the associated HTML, and the markdown file that had the entire design system. Overview and creative north star, the digital scriptorium. This design system moves away from the sterile, flat world of modern SaaS and returns to the tactile, sacred artistry of the 15th century manuscript. We are not building a website, we are crafting a living codex. So this is what it looks like to design with agents rather than just design tools. And the fact that I could with a single click bring this into the Google AI Studio experience is the value of the deep integration between these different Google AI products. Now this was a pretty easy build in large part because it got it right pretty quickly. This was the first version we saw, which was fine, but I wanted it to be closer to what Stitch had, so I copied that photo in. Along with the HTML directly, and Google AI Studio picked up and ran with it. Studio says, "I noticed that the Stitch-generated code currently uses a static image placeholder with a play button overlay instead of an actual embedded video. To make it fully functional while keeping this beautiful new design, I've taken your code and swapped out the image placeholder for the actual YouTube iframe embed. Here is your exact Stitch version ready-to-play video." Then it added, "Would you like to add any interactive JavaScript to this? For example, making the sidebar navigation functional or adding a page-turning animation?" Now a page turning animation is exactly what I had been thinking of, so we got to it. At each step of the way, Studio was both doing the work and explaining how it would work. For example, without me asking, it shared how it had made and executed a plan for mobile so that the 3D horizontal flip didn't look broken on a phone. It just switched back to a different method of turning between the pages. Now all that was left on this one was to give it the right text. So we went back to Gemini and specifically to the part of the gem where we had been brainstorming about the particular stories that we were going to focus on. And it wrote all of the companion copy that would sit alongside the videos. The last step was adding the video links and boom, we were ready to go. When it was time to push it live, for the sake of understanding Google AI Studio's full process, I went with its recommendation. Studio said, because this entire app is just one file, pure HTML, CSS, and JavaScript, you don't need a complicated server or backend. You can use static site hosting, which is completely free and takes about 30 seconds. The fast way it gave me was a drag and drop with Netlify that had the exact sequence of steps that led to basically dragging a folder from my desktop to Netlify and then boom, in literally about 15 or 20 seconds, the digital scriptorium lived. Now I will of course include links to all of these companion experiences, but you can get a feel for what we did and why it warmed the aesthetic of my late '90s computer heart. As I said though, I wanted to, in addition to that companion web experience, do something a little bit more complex as well. One of the things that's coolest about building with Google AI Studio is the ease with which you can integrate Google's AI features and tools. A game where there was some aspect of image generation and creative on-the-fly responses to scenarios seemed like a really good way to try that out. And so we dug in. We articulated the first set of game mechanics as well as some visual inspiration in Gemini, and then brought that over to Stitch. Now what you're looking at here is each of these is a different sequential generation where I was trying to get out an aesthetic that I had in my mind. So these didn't happen all at once. It was me iterating each time until I ultimately abandoned this particular canvas feeling like there was too much anchoring bias to where it had started. And with each iteration, I wasn't feeling like I was able to get it to go do something different enough. Now after a bunch of back and forth trying, I decided to try to take it in a fairly different direction just to see if there was a better way to execute what was in my mind. What was interesting is that when we landed on this style, it actually modified a little bit of the game mechanics that we had been exploring. The game that started to emerge was kind of a turn-by-turn strategy game of survival. Your goal was to live through 30 of these very volatile years, having to make decisions in a variety of difficult scenarios, using every tool in your toolbox from bribery to deceit to even actions more dramatic to survive and thrive in Renaissance Florence. From here, there was a bit of interplay between Stitch and Gemini. Where I would take something that Stitch had produced, ask Gemini to add some elements, and then ask Stitch to build those elements or something like them into the design system. And in the same way that you saw Google AI Studio being proactive and suggesting next things that it could do, Stitch does that as well. For example, the Codex page, the Diplomacy page, and the Inventory and Ledger page all came about after it said, do you want me to design the rest of these pages that we need to get the game ready to go? And once again, when we were ready, we pressed export and build with AI Studio and it brought everything in. Now this was a more complex build, but it still honestly just handled it. And what we added in Google AI Studio was not only making the thing actually work, but also integrating some of those AI tools. The two places that we built Google's AI models into this were first the generated images that come up with each scenario and story. You can see here that even though this looks like it might have been pulled from a 15th or 16th century painting, The fact that it is labeled with the same name as the scenario gives away that it was generated on the fly, and also for this game, rather than having a choose-your-own-adventure style prescriptive and limited set of scenarios, it's all generated on the fly. For example, when you discover absolute proof that the Medici family is illegally smuggling Turkish alum into Florence, thus bypassing the Pope's heavy tariffs, and you decide to return it to the Medici as a quiet gesture of loyalty, what happens next is again generated on the fly. Returning the salt-stained ledger to the Medici was met with with a heavy purse of gold and a cold nod from Piero the Unfortunate. The Medici use the records to systematically dismantle the Sforza-backed trade guilds, pushing your reputation with the ruling family to its absolute peak. However, your blatant favoritism has not gone unnoticed. The Sforza now view you as a Medici creature, and the Borgia in Rome have begun intercepting your personal couriers, sensing that the information broker of Florence has finally picked a side. Now there's a whole bunch else to this game where you can use tools, make bribes, see where you stand with different factions, It's actually a pretty fun game. But the bigger point, of course, for our purposes is that we were able to build this thing from conception to execution in honestly barely any time at all. There will be a link to the game Republic of Lies that you can go check out from the show notes. If you've been listening to the AI Daily Brief this year, you'll have heard me say a number of times that it seems to me that Google's unique opportunity is taking advantage of this wide multimodal capability and this huge array of different tools to really create differentiated and integrated capability sets. I had a ton of fun building this complete experience from concept to research to cinematic video overview to YouTube page to companion website to companion AI-driven game in the course of just a couple of hours. Hopefully this inspires you to think about what you can build, and big thanks to Google for partnering on this episode. For now, that is going to do it for today's operator's bonus edition of the AI Daily Brief. Thanks for listening or watching as always, and until next time, Peace.