Colivium helps remote workers find coliving spaces with genuine community, not just shared amenities.

Genesis

I work remotely from colivings 2-3 times per year.

The coliving market is dominated by places that are little more than house shares with strangers. Figuring out which colivings have good community is hard – many have great reviews because guests’ expectations don’t include community. You have to read between the lines of reviews to figure this out.

At some point, I realised that LLMs are pretty good at reading between the lines. So I extracted some reviews and started prompting.

Features

• Review insights: Generated from tens to hundreds of reviews of each coliving. Learn what guests appreciate, as well as what might be potential issues.

• Detailed information, consistently structured: Compare colivings more easily, and get a sense of the vibe of each place.

Implementation

Most of the information was extracted from the websites of colivings and from their reviews, using Claude Sonnet↗. It’s a hand-written pipeline that navigates websites and outputs structured data.

The website is hosted on Cloudflare Workers↗. There is no database, just JSON files that are synced to Cloudflare KV↗.