How this works

canar.ai runs on one strict rule: nothing unverified goes live. Here's the whole machine, plainly.

Where the options come from

The catalog is hand-checked: real places, verified against their own sites before going live. An AI — Claude — helps research new entries. Every spot in the catalog is read and approved by a person before you see it. Events work differently: they come from sources we've verified, they pass automated checks, and anything that turns out wrong gets pulled. And where something can't be verified — tonight's exact hours, say — the site tells you to check rather than guessing.

What's launching soon

An automated nightly pipeline that reads local event calendars and proposes additions to the catalog. Same rule applies: nothing unverified goes live.

How it learns

When you pick A or B, the site logs the matchup and your choice — anonymously. Right now that's a growing record of what Raleigh actually chooses; as it stacks up, it starts steering which options you see. Rerolls count already: telling the site both options missed means you won't see them again for a while.

What the AI doesn't do

AI writes nothing on the fly. What you read here was written or approved by a person, and the picker serves pre-written entries — there's no model improvising back there.

When it's wrong

If this site is ever wrong about hours or prices, that's a bug, and I want to hear about it. Email me: dj@wabeka.com