IdeaLens

Inspiration

The "Valley of Death" for startups isn't just a lack of funding—it's a lack of validation. Too many founders spend months building an MVP only to realize the market doesn't exist. IdeaLens was inspired by the need for a "pre-incubation" tool: a way to brutally stress-test an idea in 60 seconds before writing a single line of production code.

What it does

IdeaLens takes a raw idea and performs a multi-layered analysis:

• Scoring: Evaluates ideas on a 1-10 scale across Market Validation, Problem-Solution Fit, Business Model, and Execution.
• Competitive Intel: Maps out direct and indirect competitors, gauging market saturation.
• Validation Experiments: Generates 5 low-cost "Lean Startup" experiments to track demand.
• Simulated Community: Predicts feedback from Developers, Investors, and Customers.
• Idea Comparison: Allows founders to rank multiple ideas side-by-side to choose the winner.

How I built it

The project is built on a modern React + Vite foundation, utilizing Vanilla CSS with a custom-engineered design system for that premium glassmorphism feel.

• AI Core: Powered by Gemini 1.5 Flash with complex structured prompting to ensure high-quality JSON responses.
• Persistence: Uses localStorage for complete data privacy—your ideas stay on your machine.
• Reliability Layer: Built a custom AI service with recursive model fallbacks and a "Mock Mode" for network-restricted environments.

Challenges I ran into

• API Connectivity: We faced significant hurdles with regional model availability and network-level blocks (CORS/Firewalls). I solved this by implementing a resilient fallback chain that tries multiple model versions (v1 vs v1beta) and model types.
• JSON Parsing: Getting LLMs to consistently output valid JSON for complex nested matrices is difficult. I built a robust "cleaning" parser that uses regex to find and repair JSON structures within conversational text.
• Glassmorphism Performance: Maintaining 60fps animations with heavy backdrop filters required optimizing the CSS property stack to avoid layout thrashing.

Accomplishments that I'm proud of

• Dynamic Scoring System: The circular progress indicators aren't just static—they animate and change color dynamically based on the weighted average of the AI's logic.
• Total Privacy: Achievement of a "No-Backend" architecture where even the API keys are handled locally, ensuring user ideas are never stored on a middleman server.
• The Comparison Matrix: Creating a side-by-side ranking system that can accurately weigh the pros and cons of two completely different industries.

What I learned

I deepened my understanding of the Gemini API's versioning nuances and the critical importance of building for "Network Resilience." I also learned that math-heavy UI components (like the scoring gauges $\text{Score} = \sum \frac{w_i x_i}{n}$) require careful synchronization with React's state lifecycle to look smooth.

What's next for IdeaLens

• Web Scraping Integration: Moving from "AI knowledge" to real-time market data by scraping current Google Search and Product Hunt results.
• Pitch Deck Generator: Taking the validated data and automatically producing a 10-slide deck.
• Collaborative Mode: Allowing teams to share ideas via encrypted temporary links.

Built With

• gemini
• huggingface
• react