Turning Puddles into Oceans: Building an AI That Transforms the Mundane into Epic Tales
Building an AI That Transforms the Mundane into Epic Tales
Or how a late-night tweet led me to turn puddle-jumping into legendary sagas
It was almost midnight when I saw Alex Albert's tweet. The Anthropic team had just launched their Built with Claude Contest, and something about it felt different. Maybe it was the timing—my two-year-old was sleeping peacefully next to me, having exhausted himself jumping in every puddle he could find that afternoon. Maybe it was the challenge. Or maybe it was just that competitive itch every entrepreneur knows too well.
Three things converged in my mind: I hadn't tried Sonnet 4.5 yet. I'd been living in Cursor IDE for months and had neglected Claude Code. And most importantly, I'm pathologically competitive. Dos pájaros de un tiro, as we say back home.
But here's the thing about ambitious ideas at midnight: they rarely account for reality.
The Reality Check
My calendar was a war zone. My time is already stretched thin. Add to that an upcoming presentation at a health informatics congress in Buenos Aires this week, and you get the picture. I needed to prepare a talk on healthcare interoperability and how GenAI can accelerate standardization processes—critical topics when you're trying to connect fragmented health systems across Latin America.
(If you're in Buenos Aires this week, come say hi at the congress. Let's talk about FHIR, HL7, and how we can finally make health data work for patients, not against them.) Here you have the invitation link
So there I was: limited time, a professional obligation, and a contest deadline looming. The rational move would have been to pass. But watching my son sleep, I couldn't shake an idea that had planted itself in my brain.
The Spark: When Puddles Become Oceans
What if we could transform our daily mundane experiences into epic tales worthy of Homer? What if that afternoon puddle-jumping session could become a legendary naval battle against sea creatures? What if getting dressed in the morning could turn into a hero's quest to find the Magical Tunic of Comfort?
Kids live in a world where everything is already epic. They don't see obstacles; they see adventures. My son doesn't step in puddles—he challenges the water lords to combat. He doesn't eat breakfast—he replenishes his hero's energy for the day's quests ahead.
And that's how "Childhood Saga" (or "Epopeyas de la Infancia" in Spanish) was born. A GenAI-powered app that transforms trivial daily events into bedtime stories of mythological proportions. Because every child deserves to be the protagonist of their own epic tale.
The Token Budget Controversy
Before diving in, I hit a concerning speed bump. Twitter was buzzing with complaints about Claude Code's token consumption with Sonnet 4.5. People were burning through their weekly usage limits in hours. Some were even calling it a cynical move by Anthropic—launch a contest that forces people to upgrade their subscriptions. ROI guaranteed for them, financial pain for us.
I wasn't entirely convinced by the conspiracy theory, but the concern was valid. For an Argentine entrepreneur, $200 USD isn't just money—it's nearly a minimum wage salary. My $20 monthly subscription was already a calculated investment. I couldn't afford to accidentally blow through hundreds of dollars in API costs while building a contest entry.
So I did what any good engineer does: I took screenshots. Before writing a single line of code, I documented my starting usage metrics. This wasn't just about building an app anymore; it was about testing whether the democratization of AI development is real or just another premium service disguised as accessibility.
Time to find out.
The Setup: Tools, Constraints, and First Moves
I spent the first hour getting Claude Code properly configured on my machine. Fresh install, clean slate. Then came the crucial part: crafting the initial prompt. Not just any prompt, but one that would capture the essence of what I wanted to build while giving Claude Code enough context to architect something meaningful.
The concept was clear in my head:
- Parents input simple daily activities
- The app transforms them into epic narrative frameworks
- Stories maintain the child's perspective as the hero
- Multiple story modes: adventure, mystery, comedy, legend
- Bilingual support because my son is growing up between cultures
But translating that vision into a prompt that Claude Code could work with? That required thinking about structure, constraints, and the narrative voice I wanted to achieve. The stories needed to be epic but not scary. Empowering but not condescending. Imaginative but grounded in the actual events of the day.
I opened a new file, stared at the cursor blinking, and started typing. Not code—at least not yet. First, the blueprint. The architecture of wonder.
And then, finally, I gave Claude Code the go-ahead.
In the next chapter, we'll dive into the actual development process: what worked, what didn't, how many tokens I actually burned, and whether Claude Code lived up to the hype—or the fears. Stay tuned for the play-by-play of building with Sonnet 4.5 under pressure.
Because if we're going to talk about democratizing technology, we need to be honest about what it actually costs—in time, money, and tokens.
Next in this series:
- Chapter 1: Turning Puddles into Oceans: Building an AI That Transforms the Mundane into Epic Tales
- Chapter 2: The Development Sprint
Built from Buenos Aires with mate, determination, and the belief that technology should serve stories worth telling.