hotdog tier list 🌭

crawl a simcluster, read everyone's hot dog posts, rank the toppings S to F

a handle, @handle, DID, or bsky.app profile URL —

bring your own vision-model key

Optional. CLIP picks out the photos that look like hot dogs, then each of those is sent to a vision model with your key, which names the toppings and feeds them into the ranking. The key is stored in this browser and sent only to the provider you pick, never to this site.

· · get a key

Jev isn't on the list: its API sends no CORS headers, so a browser page can't call it.

how it works: for every mutual, their own post history (text + image alt text) is searched for hot dog, hotdog, chili dog and corn dog. Toppings named in those posts are counted by how many different people mention them; "no ketchup", "gross" and friends push a topping toward F. With a vision-model key saved, the photos CLIP thinks are hot dogs go to that model instead and its topping list joins the ranking. Tick "also read the photos": a quick colour check drops images with no bun, sausage or mustard tones, then CLIP (running in your browser) looks at every image that remains each mutual has posted, keeps the ones that look like a hot dog even when the post never says so, and takes a guess at toppings in them; those show up marked as spotted-in-photo. That is every photo they've posted, so it can take a while — the stop button keeps what's found so far. It's zero-shot similarity, so it's good at mustard-on-a-dog and shaky on small garnishes.