Discover.io
Cuts hours of 'which AI tool do I even need' research down to one guided conversation, with zero hallucinated recommendations.
The AI tools market is too crowded and changes too fast for a generic search engine to answer 'what should I use for this' honestly. A wrong recommendation costs a professional real time evaluating a tool that was never going to fit their workflow. Discover.io solves that with a guided clarification flow that extracts the user's actual persona, task, and success criteria before recommending anything, then ranks results from a verified tool database instead of letting an LLM invent plausible-sounding options that don't exist. The extra 30 seconds of clarification is a deliberate trade: a small amount of user patience up front, in exchange for a recommendation worth acting on.
The frontend is a Client-Side Rendered Single Page Application using React and Vite. TanStack React Query manages all asynchronous server state with efficient data caching, automatic retries, and seamless backend synchronisation. API client logic is decoupled from UI components via custom hooks built on Axios. The multi-step AI search flow (Input → Clarification → Confirmation → Results) is managed locally, leveraging Framer Motion for perceived zero-latency transitions between steps.
State Management
Chose
TanStack React Query for server state
Why
It provides built-in caching, background refetching, and robust error handling, which is crucial for managing variable response times from AI endpoints and preventing redundant network requests.
Trade-off
Requires stricter structuring of query keys and mutations compared to standard useEffect data fetching.
UX Flow
Chose
Multi-step 'Clarification' flow before executing AI searches
Why
AI recommendation engines often misinterpret generic queries. By explicitly extracting the User Persona, Core Task, and Success Criteria — and requiring user confirmation — the accuracy of recommendations improved drastically.
Trade-off
Added an extra confirmation step to the user journey in exchange for significantly higher-quality, context-driven results.
Build Tool
Chose
Vite over a full-stack framework like Next.js
Why
Vite provided lightning-fast HMR and optimised build times for an MVP that is strictly a Client-Side Rendered application interacting with external AI APIs.
Trade-off
Sacrificed out-of-the-box SEO optimisation and server-side routing capabilities, which may necessitate a migration if organic SEO becomes a primary growth vector.
Migrate to Next.js for Server-Side Rendering to improve public SEO indexing of the skills and tools catalog
Implement streaming responses (Server-Sent Events) for AI interactions to further reduce perceived latency during the Diagnosis phase
Introduce persistent user profiles to cache onboarding data for long-term personalised recommendations