How this site is built

All data comes from Thirdwatch's Product Hunt scraper on Apify. It walks Product Hunt's daily leaderboard page (/leaderboard/daily/YYYY/M/D) and extracts the ranked top-10 products for any date range you give it.

The initial 24-month scrape (April 2024 → today) took 32 minutes and cost $0.27 on Apify's pay-per-event pricing. A small products-mode run (to resolve website URLs) added another $2.50. Total end-to-end data pipeline cost: under $3.

You can reproduce all of this yourself. The scraper accepts startDate / endDate / maxPerDay inputs and takes a few lines of curl to run.

After scraping, we HEAD-checked every product's website URL. 342 returned as "dead." But a naive HEAD-check is a bad proxy for company death — it flags every rebrand, every stale marketing URL, every acquired-and-redirected domain.

So we did a second pass: for the top 100 dead by launch-day votes, a research agent did a web search on each to classify what actually happened. That gave us the audited numbers you see on /graveyard.

Next.js 16 (App Router, Turbopack, React 19) + Tailwind v4 + Recharts, deployed on Railway. The chat endpoint proxies OpenRouter with a Qwen model — cheap, accurate, fast. A single compact fact-sheet about the dataset gets stuffed into the system prompt so the model can answer questions without an expensive vector DB.

Data updates are just a daily cron: re-run the scraper for yesterday's leaderboard, append to the JSON, redeploy. The static-JSON-in-public-folder approach keeps the whole thing under $5/month.

This site is both a product and a demo. If the insights we surface made you curious about the data, the scraper is right there: you can point it at any date range, scrape into your own dataset, and ask your own questions.

We'd rather have 10 users running the scraper for their own analyses than 10,000 users staring at ours. The site is the marketing.