InfiniteGame

• 

  Campaign Breakdown

• 

  Campaign View

• 

  Mission View

• 

  Recording your inputs

• 

  Main Screen

💡 Inspiration

I’ve always struggled with turning my goals into action. I would write down dreams, set intentions, try productivity apps — but none of them helped with the real problem: starting. I realized what I needed wasn’t a better app — it was a better conversation. Something (or someone) that could listen, help me unpack my goals, break them into steps, and keep me emotionally invested.

The idea for InfiniteGame came from this gap — the longing for a guide.

A character. A coach. A friend.

Not just a to-do list, but a dialogue with your future self.

🧠 What it does

InfiniteGame is an MVP for an AI to-do app powered by conversation.

• You speak your goals aloud to a lifelike character.
• The character responds with an encouraging message and breaks your goals into actionable tasks using GPT-4.
• These tasks are scheduled and shown in a task list with a dynamic calendar view.
• You can view, complete, and manage your to-dos in an intuitive, gamified interface.

It’s not just productivity — it’s emotional momentum.

🛠️ How I built it

• Frontend: Built with Vite/React, Tailwind CSS, and shadcn/ui for fast, reactive UI.
• Speech Recognition: ElevenLabs for capturing user input via microphone and performing Speech-To-Text.
• AI Logic: OpenAI GPT-4.1 processes speech and returns two outputs: a friendly spoken message and a structured JSON with tasks.
• Speech Generation: ElevenLabs performs Text-To-Speech and responds to the user.
• Backend: Supabase handles authentication, task storage, and real-time updates.
• Deployment: Deployed to Netlify with Bolt's integration.

⚔️ Challenges I ran into

• Overthinking and paralysis: I spent weeks thinking through every option — from animation pipelines to hosting — and almost never started.
• Time crunch: Although I ideated on May 30th, I only started building 2 days before the deadline. That made shipping even a basic MVP a sprint against time.
• Stack overload: Tools like ElectricSQL were evaluated and dropped to keep the scope realistic. Also I discovered that Remix, a framework I was planning on using, upon request was actually transformed into Next.js, however Next.js didn't have support for Supabase integration in Bolt.new, so remaining with the standard Vite/React/Typescript starter was essential
• Overly complicated MVP: The initial MVP was supposed to use a 3D avatar created with Ready Player Me and Matomo for animating the character and using a Text-To-Speech from Azure that provided viseme data, however, this proved to be way too complicated to implement, so I remained with a simple pulsating orb that did the talking
• Working with Bolt.new: In general it's quite a tedious process of going back and forth and sometimes debugging why something doesn't build was quite complicated.

🏆 Accomplishments that I'm proud of

• Built a working full-stack MVP for an AI-powered assistant in under 48 hours.
• Seamlessly combined cutting-edge voice tech and LLMs into a single experience.
• Created a meaningful emotional arc — not just a product, but a presence that people connect with.
• Stayed focused on the experience, not just the features.

📚 What I learned

• Start before you’re ready — clarity comes from doing, not thinking.
• It’s better to have one feature that feels magical than ten that feel meh.
• LLMs are great at planning when you give them structure (JSON outputs, tagged blocks).
• People respond to emotion a lot and it's probably one of the main things that influence us, we'd like to think we're rational, but in fact people are highly emotional

🚀 What’s next for InfiniteGame

• First of all implementing a subscription plan for the product as it is currently running on the ElevenLabs Creator Plan and also using my $100 OpenAI credit for generating everything and it can become costly quite fast
• Implementing the original vision for the product:
  • 3D Avatar: Created using Ready Player Me, animated with Mixamo, and rendered in Three.js.
  • The avatar lip-syncs your assistant’s voice in real-time and celebrates when you complete tasks.
  • Lip Sync & Voice: Azure Text-to-Speech with viseme support for real-time lip-syncing.
• Improve the overall To-Do app experience and make it more robust, currently still in MVP phase
• Add memory: let the assistant learn from your patterns and past completions.
• Improve character emotional range and facial animations.
• Allow natural edits to tasks via voice (“Move that to tomorrow”).
• Add more celebration animations and streak tracking.
• Explore mobile and AR versions of the experience. Currently only the desktop version is optimised.
• Long-term: let the assistant evolve with you — as a co-pilot for personal growth, not just productivity.

Challenge Compliance

• Deploy Challenge (Use Netlify to deploy your full-stack Bolt.new application): I have successfully deployed InfiniteGame as a full-stack application using Netlify, the provided link showcases my ability to deploy to Netlify: https://sage-twilight-13f7da.netlify.app/
• Custom Domain Challenge (Use Entri to get an IONOS Domain Name and publish your Bolt.new app on the domain): Bought the custom domain infinitegame.site for the project. The name chosen for the project is inspired partially from Simon Sinek's book The Infinite Game and I see self-improvement as one of those games that you can play to infinity.
• Voice AI Challenge (Use ElevenLabs to make your Bolt.new app conversational): The voice function is the main driver of the app and I've used the Jessica voice from ElevenLabs for all interactions. Also I used ElevenLabs for the Speech-To-Text
• Startup Challenge (Use Supabase to prep your Bolt.new project to scale to millions): I've used Supabase as the Database, Auth and Storage provider for the project

Built With

• bolt.new
• elevenlabs
• netlify
• radixui
• react
• shadcn-ui
• supabase
• vite