Inspiration

The inspiration for ChainPulse stems from a major pain point in the current Web3 PM landscape: Tool Fatigue.

Currently, Web3 PMs are forced to navigate a fragmented ecosystem, jumping between multiple tools for on-chain analytics, governance monitoring, social sentiment, and smart contract audits. For someone transitioning into a Web3 PM role, the learning curve is incredibly steep. I wanted to build a single "Strategic Hub" that lowers this barrier to entry and consolidates day-to-day productivity into one unified intelligence layer.

What it does

ChainPulse is a world-class AI-powered product intelligence engine designed specifically for Web3 Product Managers. It synthesizes on-chain signals, market sentiment, and competitive benchmarks into a comprehensive "13-Pillar" strategic view, transforming fragmented data into actionable product roadmaps.

How we built it

The development of ChainPulse was a structured journey that prioritized planning and documentation as much as the code itself. The process involved:

  1. Planning and Roadmap: Defined the "13 Pillars" and the core MLP (Minimum Loveable Product) requirements.
  2. Iteration and Grounding: Refined the AI strategy, specifically integrating Google Search Grounding to ensure the data was live and relevant.
  3. Application Development: Built the React 19 frontend and integrated the Gemini API.
  4. Documentation and Context Logging: Developed a PRD, TRD, and Roadmap to keep the project's vision aligned.
  5. GitHub Versioning: Connected my GitHub account to manage versions and ensure a clean deployment pipeline.
  6. Feature Augmentation: Added deep-dive capabilities for GitHub URLs, X handles, and smart contract addresses.
  7. Rigorous Debugging: Tested features to ensure consistent AI output and stable UI performance.
  8. Continuous Documentation Updates: Logged every new idea and progress milestone back into the project docs to maintain a "living" context for the AI.

Challenges we ran into

The development was not without its hurdles:

  1. LLM Hallucinations: Managing the AI's tendency to invent data required strict prompt constraints and grounding verification.
  2. Debugging Complexity: Debugging AI-generated errors and maintaining state across 13 different modules took longer than initially anticipated.
  3. Iteration Intensity: The project required many more iterations than a traditional software project to ensure the "PM logic" in the reports was truly world-class.

Accomplishments that we're proud of

  1. Strategic Pillar Synthesis: We successfully engineered a system that consistently maps disparate Web3 signals into 13 high-impact capability pillars without data loss.
  2. Real-Time Grounding Precision: The integration of live search grounding effectively bypassed the "training data cutoff" problem, providing insights on events that happened only hours ago.
  3. Boardroom-Ready Reporting: The HTML export engine produces reports that aren't just data dumps, but professional strategic documents ready for DAO stakeholders.
  4. Privacy-First Architecture: Building a powerful AI hub that is 100% client-side, ensuring user API keys and protocol strategies never leave their local environment.

What we learned

Through the process of "Vibe Coding," I realized that Prompt Engineering is only the beginning.

  1. Context Engineering is King: The quality of the AI's output depends entirely on the richness of the context provided (project parameters, competitors, and specific data URLs).
  2. Iteration is Non-Negotiable: A good product isn't built in one shot. It requires constant feedback loops between the developer and the LLM to refine logic and UI.
  3. Planning Prevents Hallucination: Before starting any development, creating a solid plan and roadmap is essential. Without a roadmap, it is easy to lose track of the product's core purpose.

What's next for ChainPulse

  1. Direct API Ingestion: Integrating native connectors for the GitHub API and Etherscan to supplement qualitative AI findings with hard technical metrics.
  2. Social Pulse Expansion: Developing a deep-sentiment layer that monitors X (Twitter) and Farcaster in real-time to track community loyalty shifts.
  3. AI-Drafted Governance: A "One-Click Proposal" feature that turns "Actions Required" findings into ready-to-post Snapshot or Tally governance proposals.
  4. Predictive Risk Modeling: Moving from reactive intelligence to proactive modeling for liquidity crunch and de-pegging events.

Built With

  • browser-localstorage
  • css
  • environment-variables
  • gemini-3-flash
  • gemini-3-pro
  • github
  • github-jobs
  • google-cloud-run
  • google-search-grounding
  • html5
  • lucide-react
  • markdown
  • npm
  • react-19
  • tailwind-css
  • typescript-es6+
  • vite
Share this project:

Updates