TL;DR — I built Watch Arbitrage Tracker ( Apify Store , GitHub ): a Crawlee + Camoufox actor that scrapes 6 luxury-watch marketplaces in parallel, computes the cross-platform median price for any Patek/Rolex/AP reference, and pings Telegram the moment a listing drops more than X% below market. Sub-$1/month for typical dealer usage. Doubles as an MCP server so Claude Desktop / Cursor / ChatGPT can query the live feed in plain English. The interesting part isn't the build — it's the 5 bugs I had to debug in production after pushing public , and the cross-platform median math that turns "scraped data" into a real arbitrage signal. The problem (real, validated, painful) Pro watch dealers — the people who flip pre-owned Patek 5711, Rolex Daytona, AP Royal Oak — spend 3+ hours a day refreshing 6 dealer marketplaces looking for mispriced inventory. The job is mechanical: open Chrono24, search 10 reference numbers, compare prices, switch to WatchBox, repeat, switch to Bobs Watches, repeat...…