Promptle – The Daily AI Prompt Challenge

🚀 My Hackathon Build: Promptle – The Daily AI Prompt Challenge

🌟 Inspiration

Ever feel stuck trying to get the right output from ChatGPT or Midjourney? The idea sparked from a simple thought: “What if Wordle met ChatGPT?” I wanted to gamify the creative process of writing AI prompts—a skill becoming as vital as coding. As an avid prompt tinkerer and daily Wordle player, I felt there was a missing intersection between creative constraint, feedback, and skill-building. Thus, Promptle was born: a daily game where players write prompts instead of guesses, and creativity—not just logic—is the path to victory.

Promptle is a fun, Wordle-style game that helps you learn to write better AI prompts—in just a few minutes a day. It’s for creators, builders, and anyone who wants to level up their prompting skills.

This is built for mobile and desktop web page use for people who want to do this on their phones on the go, and for those who want to do it at their desk, they have access to the web app as well.

🎮 What It Does

Promptle is a daily game where you don’t guess words—you write prompts.

Each day, players are given a creative challenge, like:

“Describe a city that only appears during eclipses”
“Write a dialogue between a dragon and a time traveler”
“Generate a recipe for an alien smoothie”

You write a prompt and get scored on accuracy, creativity, and efficiency. You have 3 tries—and hints help you get closer. If you nail it, you unlock a bonus round with a twist, like ‘Now rewrite it as a limerick!’

Your mission is to write a prompt (for ChatGPT, DALL·E, etc.) that gets the AI to produce an output as close as possible to a hidden, pre-defined AI-generated ideal response.

Core Mechanics

3 Tries Per Day🧠 : Refine your prompt with feedback after each attempt
AI Scoring🎯 : GPT evaluates your prompt based on:
Accuracy: Did your output match the intent?
Creativity: Was your idea fresh, fun, or vivid?
Efficiency: Was your prompt short, clear, and effective?
Color-Coded Feedback🟩 : Just like Wordle (green = great, yellow = okay, red = off)
Bonus Rounds🔓 : Nail a perfect 10/10 or 9/10 and unlock fun bonus challenges like: “Rewrite it as a poem” “Turn it into a noir detective story”
Streak Tracking🔥 : Keep your creativity flowing day after day
Analytics and events tracking to help me know local stats for the day and how many users used it, how many bonus rounds were unlocked, what prompts were sent to score a user prompt, etc.

All of this happens client-side, with no accounts or sign-ins—just one click and you’re playing.

It’s a mini gym for your prompt skills—with a daily dose of creative joy.

🛠️ How I built it

🧩 Built Entirely in Bolt

This entire project was created and iterated inside Bolt.new, a zero-setup IDE and playground. Why?

It let me rapidly prototype frontend logic in React + TypeScript
I could test OpenAI prompts, refine UI flows, and even handle local storage-based streak tracking—all in-browser
No need for local dev environments. Every update was live, every tweak instantaneous.

🧱 Core Tech Stack

Layer	Stack
Frontend	React 18, Tailwind CSS, TypeScript
Animations	Framer Motion
AI Layer	OpenAI (GPT-4 Turbo, GPT-3.5)
Scoring	Custom GPT evaluator + similarity
Storage	LocalStorage (for streaks/stats)
Audio	in built since ElevenLAB did not work for my use
Deployment	Netlify

⚙️ Key Features done in Bolt directly

Daily challenge prompt system with 24-hour reset logic
Score computation using OpenAI + embedded “ideal prompt” comparisons
3-try system with colour-coded feedback (for scoring with medals)
Bonus Rounds triggered by perfect scores
Local streak and score storage with celebratory popups
Shareable results via clipboard (Wordle-style but with a twist)

🧪 Prompt Scoring Logic

Embedding-based similarity scoring didn’t always capture tone or creativity
I had to fine-tune how GPT-4 judged prompts and added a handcrafted rubric to balance it out
Writing meta-prompts for GPT to evaluate user prompts felt like teaching a judge to be fair and fun

🔐 OpenAI API Security in Bolt/Netlify

Since Bolt is client-side and Netlify deploys are static, storing the API key securely was tricky
I used .env variables for Netlify and demoed a non-AI fallback mode that uses mock scoring (so anyone can play!)

🎤 Challenges I ran into

😵 Eleven Labs Voice Integration

I wanted Promptle to speak your score results back—“You got an 8/10!” with some sass. But:

Voice API limits meant I had to cache responses cleverly to avoid burning tokens
Playback inconsistencies in mobile browsers made testing a pain

🧠 What I learned

Accomplishments that I am proud of

Prompt Engineering as Game Design: I explored how scoring systems could gamify writing prompts, making users not just better writers, but better thinkers.
AI as a Judge: I learned how to leverage OpenAI’s GPT-4 and 3.5 Turbo for evaluating creative input with human-like nuance, scoring for accuracy, creativity, and efficiency.
Embedding Feedback Loops: Designing intelligent hints and bonus rounds taught me a lot about how feedback encourages iteration.
Audio Feedback (attempted using Eleven Labs): I learned how to add immersive audio using voice synthesis, though not without a few hiccups (more on that below!).
🚀 Deployed to Netlify in <5 min: Build for production, npm run build

🥇 Impact

A friend of mine used to write really vague prompts. After a week of Promptle, she learned to be super specific—like adding tone and structure. Another friend learned how to write image prompts for DALL·E that actually worked. They both said Promptle made prompting feel like a game—and now they use those skills at work.

What's next for Promptle

Creative Prompt Challenge Game
Multimodal Prompts: Let users write for DALL·E, Eleven Labs, or even video
Social Remixing: See others’ prompts and remix them
Hard Mode: Start with just a title or emoji (e.g. “🌌🧃” = “Create a cosmic smoothie recipe”)
Leaderboard: Weekly themes, top scorers, and global streaks