November Rain & The AI Explosion
We survived the AI plateau. Now we are Vibe Coding. A look inside the engine room of Mythoria and the wild new models released this week.
Last weekend it was raining in Porto. The kind of rain that makes you want to curl up with a good book—or in my case, curl up with a terminal window and about fifty tabs open in my browser. ☔
If you’ve been following the tech world this past week (November 18th to 25th, 2025), you know that the "AI Winter" didn't just end; it evaporated. It feels like every major lab decided to drop their mic at the exact same time. We got Google’s Gemini 3, OpenAI’s GPT-5.1, Claude Opus 4.5, and xAI’s Grok 4.1.
For a startup founder in his early 20s trying to build the world’s best personalized book platform, this is both a blessing and a "how do I integrate all of this before I sleep?" nightmare.
Here is a look inside my head—and the Mythoria backend—on how these crazy new tools are rewriting our story.
The Age of "Vibe Coding" 👨💻
First, let’s talk about building the ship. You know that feature on our backlog? The one to improve the PDF generation for printing? It’s been sitting there for months because handling that massive codebase was... painful.
Enter Claude Opus 4.5 and GPT-5.1-Codex-Max.
We call it "Vibe Coding" now. I don't just write syntax anymore; I tell the AI the intent, the feeling, the architecture, and it acts like a senior engineer.
I sat down with Claude Opus 4.5 yesterday to finally tackle that print generation module. In the past, the model would lose context halfway through the file. It would forget we were using A5 paper size or hallucinate a library we didn't have. But Opus 4.5? It held the entire context. It refactored the whole thing in one session.
It feels less like coding and more like conducting an orchestra. 🎻
The Storytelling Brains: Gemini 3 vs. The World 🧠
At the heart of Mythoria is the story. We take your memories and turn them into magic.
We’ve been running tests all week replacing our old reliable Gemini Pro 2.5 with the shiny new Gemini 3 (and its "Deep Think" capability).
The Verdict: The stories are undeniably better. With Gemini 2.5, sometimes the plot would drift. A character might forget they were holding a magic sword in Chapter 3. With Gemini 3, the causality is tight. The twists feel earned, not random. It plans the narrative arc before writing the first word.
The Dilemma: It’s expensive. 💸 The price hike for these "reasoning" models is real. Does a personalized bedtime story need the cognitive power of a PhD student? Maybe not always. We are finding that Grok 4.1 actually has incredible "EQ" (Emotional Intelligence). It writes dialogue that feels warmer, more human, perfect for kids' books.
So, the future of Mythoria isn't just one brain. It’s a router. We might use Grok for the heartwarming hug, and Gemini 3 to make sure the facts about the Solar System in your educational book are 100% accurate.
A Quick Note on Names (Are you listening, Sam?) 😅
I have to make a joke here. OpenAI, I love you, but your naming scheme is a tragedy.
Sam Altman said he wanted to simplify the product line with GPT-5. But now we have:
- GPT-5
- GPT-5.1
- GPT-5.1-Codex
- GPT-5.1-Codex-Max
It’s getting harder to choose a model than it is to choose a restaurant in downtown Porto on a Saturday night. It’s a menu with too many pages!
Visuals: The "Nano Banana" Situation 🍌
Yes, you read that right. Google released a state-of-the-art image model and called it Nano Banana Pro. Apparently, it was a codename that stuck.
We tested it against our current OpenAI image generator. The results? Banana wins on consistency.
The biggest pain point in AI art is that "Character A" looks different on Page 1 than on Page 5. Nano Banana Pro fixes this. You can upload a reference sheet, and it respects the geometry of the face perfectly.
However... For now, we are sticking with our current setup. Why? Cost. Nano Banana is brilliant, but it's premium. We want Mythoria to remain accessible to every family, not just a luxury item. But mark my words: as soon as that price comes down, or we launch a "Mythoria Pro" tier, the Banana is coming.
Where are we going? 🚀
If this week proved anything, it's that the speed of innovation isn't slowing down. It’s accelerating.
I’m looking at Veo 3.1 (Google’s video model) and Sora 2 (OpenAI video model) and dreaming. How long until your Mythoria book comes with a 30-second movie trailer? You scan the QR code on the back of your printed book, and boom—your child is starring in a movie trailer for their own adventure.
Technically? We could build a prototype today. Economically? Give it 6 months.
But for tonight? Tonight I have a smaller battle. I’m going to spend the next few hours tuning our Text-to-Speech engine. The new audio models are great, but they still get confused between European Portuguese and Brazilian Portuguese accents. And nothing ruins a story about the Douro River like an accent from Rio de Janeiro! 🇵🇹
Back to the code mines. Or rather, back to the Vibe.
Keep dreaming, keep reading.