Inspiration Product managers live on stats, headlines, and hot takes — in decks, Slack threads, and LinkedIn feeds. I kept seeing the same problem: someone shares a bold claim (“90% of startups fail,” “Vitamin C prevents colds,” “AI will replace PMs”) and the room nods without checking.
Verifying that takes opening many tabs, guessing which sources matter, and still not knowing if the claim is fair. I built Clarify for Mind the Product’s “Everyone Ships Now” challenge: a small product that does one job well — paste a claim, get a clear verdict with trusted sources in seconds.
What it does Clarify is a web app for PMs, designers, and curious people who want a fast second opinion before they share something.
Paste any headline, stat, or health claim Check — Clarify normalizes the claim and cross-checks against high-trust sources (WHO, CDC, Reuters, HBR, and similar) Get a verdict — Supported, Mixed, or Disputed, with confidence and cited sources plus trust scores No chatbot rabbit holes. One input, one structured answer, ready for a deck, doc, or team thread.
Live app: https://verdict-validator-tool.lovable.app
How we built it I shipped end-to-end in under a month with an AI-first stack:
Layer Tool Frontend Lovable — polished two-panel UI (input + verdict) Backend Python + FastAPI — claim extraction, source scoring, verdict logic Deploy Render for the API; Lovable for the public product URL Development Cursor for rapid iteration on engine and API Analytics Novus.ai — product instrumentation and dashboard for hackathon eligibility The engine runs a simple pipeline: Extract → Research → Verdict, with trust-weighted sources so low-quality sites don’t drive the answer.
Repo: https://github.com/prakash023-hub/clarify-
Challenges we ran into Shipping beats polishing — Early Gradio prototypes worked but didn’t feel like a product; moving the UI to Lovable while keeping Python as the brain took careful API design and CORS setup. Backend vs frontend split — Judges need a clickable URL, not “clone and run locally.” Wiring Lovable to a live Render API and testing cold starts was essential. Source quality — Open web search is noisy; scoring and filtering to trusted domains was more important than adding more features. Demo reliability — With optional live search, I built a demo mode with curated high-trust fallbacks so every demo claim returns a useful verdict on submission day. Accomplishments that we're proud of Shipped a real product — Strangers can land on the URL and check a claim without setup. Clear product sense — One user (PMs), one job (sanity-check claims), one flow (paste → verdict). Full stack solo — Idea → Lovable UI → FastAPI engine → Render deploy → Novus install. Works in the demo — Health and business claims (e.g. startup failure stats, vitamin C and colds) return structured verdicts with sources judges can follow. What we learned Mind the Product’s “Everyone Ships Now” framing is right: a small, working product beats a big slide deck.
I learned to:
Define scope tight enough to ship in weeks Put the live URL first and treat the repo as support Design for judges and users in 30 seconds — clear headline, one button, obvious output Use AI builders (Lovable, Cursor) for speed, but keep one custom engine where the product logic lives Shipping Clarify in public also reinforced that product work is as much about what you leave out as what you build.
What's next for Clarify — Mind the Product World Product Day 2026 Live search by default — Full Nimble-powered web research for every query Browser extension — Highlight text on any page → instant Clarify verdict Slack agent — Fact-check claims where teams already talk (workspace integration in progress) Team mode — Shared history of checked claims for PM and research teams Stronger PM workflows — Export verdict + sources as a one-slide “evidence snippet” for roadmaps and stakeholder updates Clarify started as a hackathon ship; the goal is to keep it small, trustworthy, and fast — the tool I wish existed before every roadmap review.
Built With
- fastapi
- httpx
- lovable
- novus.ai
- pydantic
- python
- react
- render
- typescript
- uvicorn
Log in or sign up for Devpost to join the conversation.