Hunting for a place near UBC meant checking five-plus sites, Wesbrook, liv.rent, Craigslist, apartments.com, Zumper, every single day. That's 200+ searches, and most of what comes back is junk. Worse, the one thing you actually filter on, the move-in date, is almost never a field. It's buried in a human-written blurb ('available Sept 1, 2 bed, no pets…'), so the listings for the month you want get lost under everything else. A lot of wasted time.

Describe what you want in plain English and an LLM turns it into a search. It pulls listings from all the sources at once, reads the real move-in date out of the free text, scores each against your criteria, flags likely scams, and ranks the survivors in a local dashboard you can filter by beds, price, distance, and move-in window. New listings get a badge; ones that vanish are marked gone, not deleted.

The interesting part was that every site exposes its data differently, so the effort escalates per source. liv.rent is a server-rendered Next.js site with the full listing embedded as JSON in the HTML, so I just pull that out, clean and structured, no LLM. Craigslist has no structure at all, just a paragraph a human wrote, so that's where the LLM earns its place, extracting the move-in date and fields from prose. apartments.com and Zumper actively block scrapers, so those run through pre-built Apify actors that handle the anti-bot layer. Same goal every time, structured listings, with escalating effort depending on how the site hides its data.

I found my current rental on the first day I used it.