Project Story: Polkadot Discovery

Inspiration

As a first-time hackathon participant, I was drawn to the Polkadot Hackathon’s challenge to innovate in Web3, a paradigm I see as a solution to the biases and gatekeeping of centralized platforms. Reading the hackathon description on November 11, 2025, I reflected on how platforms like Google prioritize paid content, marginalizing small businesses and creators. This inspired Polkadot Discovery, a decentralized content discovery platform built on Polkadot’s Westend Testnet to empower users and creators with transparent, bias-free search.

My vision was to address three Web3 challenges:

1. Countering Centralized Bias: Create a search system where discoverability relies on relevance, not ad budgets, using blockchain-based tagging.
2. Enhancing User Experience: Leverage machine learning (ML) for dynamic, topic-based search with user-controlled preferences.
3. Respecting User Attention: Deliver curated, non-intrusive newsletters instead of spammy notifications.

Web3, to me, is about sovereignty and fairness. Polkadot’s interoperable blockchain offered the perfect foundation to prototype a system where users own their data and creators gain visibility without centralized intermediaries.

What I Learned

This hackathon was a crash course in Web3 development and collaboration with large language models (LLMs). Key lessons include:

• LLM-Driven Development: I honed a process of iteratively prompting LLMs to design and code, translating my Web3 vision into a functional PoC. This required clear communication to align LLM outputs with my SCIS framework.
• Polkadot Ecosystem: Integrating Polkadot.js and the Westend Testnet taught me about blockchain wallet connections and on-chain data storage (e.g., system.remark extrinsics for preferences).
• Decentralized Architecture: I learned how to structure a flat-file system (SCIS) for traceability, with 31 files organized as p{phase}s{section}f{file}.{extension} (e.g., p0s0f1.js for the orchestrator).
• Time Management: Building a PoC in ~6 days pushed me to prioritize features like dynamic filtering and newsletter previews while planning for future scalability (e.g., #tagging system).

How I Built It

Starting November 11, I used my LLM process to design and build Polkadot Discovery in under a week. Here’s the approach:

1. Ideation with LLMs:

   • I prompted LLMs to define Web3 challenges, focusing on Polkadot’s scalability and user sovereignty. This shaped the PoC’s three pillars: decentralized search, ML-enhanced search, and newsletters.
   • I outlined a system to search across topics (movies, restaurants, travel, events, tech) with dynamic filters and on-chain preferences.

2. SCIS Organization:

   • I applied my Sequential Component Identification System (SCIS) to structure 31 files in a flat directory (C:\Dev\Polkadot). Files follow p{phase}s{section}f{file}.{extension} (e.g., p1s0f6.jsx for the React app component).
   • Components are named Cp{phase}s{section}f{file}.{index} (e.g., Cp0s0f4.1 for the query analyzer). This ensured traceability and LLM compatibility.

3. Tech Stack and Implementation:

   • Frontend: React (^18.2.0), Vite (^5.4.21), Tailwind CSS (^3.3.0) for a responsive UI, served at http://localhost:5173/p1s0f1.html.
   • Backend: Node.js (v16+), Polkadot.js for wallet integration, and Westend Testnet for preference storage.
   • Data: JSON mock datasets (p1s1f1-5.json) for content and topic mappings (p1s1f6.json).
   • ML: A keyword-matching algorithm (p0s0f4.js) for topic detection, with plans for advanced models.
   • Tools: ESLint, Prettier, and @vitejs/plugin-react (^4.0.0) for code quality.

4. Feature Development:

   • Search and Filters: Users query topics (e.g., “thriller movies 2024”), and ML detects topics to load filters (p1s0f3.jsx) from p1s1f6.json.
   • Newsletter: Curated updates (p1s0f4.jsx) based on filters, avoiding notification overload.
   • Polkadot Integration: Wallet connection (p1s0f5.jsx) saves preferences on-chain (p0s0f6.js).
   • Future Vision: Planned a #tagging system (e.g., #VeganEats) for creators to boost visibility, stored on-chain.

5. LLM Coding:

   • LLMs generated code for 31 files (9 JSON, 5 Markdown, 7 JSX, 1 CSS, 1 HTML, 12 JavaScript), guided by SCIS and my prompts. I reviewed and refined outputs to ensure functionality.

Challenges Faced

Building Polkadot Discovery was exhilarating but not without hurdles:

• Time Constraints: With ~6 days from November 11 to 17, I had to scope features tightly. I prioritized core functionality (search, filters, newsletter, wallet) over advanced ML or real APIs, deferring them to future plans.
• Web3 Learning Curve: As a first-time Web3 developer, integrating Polkadot.js and Westend Testnet required rapid learning. Debugging wallet connections (p1s0f5.jsx) was tricky but rewarding.
• LLM Alignment: Ensuring LLMs produced SCIS-compliant code demanded precise prompts. I iterated to align outputs with p{phase}s{section}f{file} naming and component logic.
• Styling: Configuring Tailwind CSS (p1s0f7.css) for a polished UI was challenging under time pressure, requiring manual tweaks.

Looking Forward

Polkadot Discovery showcases Web3’s potential to democratize content discovery. The #tagging system, planned for future iterations, will let creators tag content (e.g., #IndieFilm) on-chain, enhancing visibility without centralized bias. This hackathon taught me the power of LLMs and Polkadot to build user-centric solutions. I’m excited to expand this PoC with real APIs, advanced ML, and a Polkadot parachain to create a fully decentralized discovery layer for the internet.

Built for the Polkadot Hackathon, November 2025
Polkadot Discovery Team

Built With

• claude
• css
• gemini
• grok
• javascript
• perplexity
• scis-(sequential-component-identification-system)