Jasper
@raedisch.net
Thanks for the feedback!
November 20, 2025 at 11:14 AM
Thanks for the feedback!
It matches anyone that is indexed to the listed accounts. The scale is a little random, only based on my matches (my worst cosine similarity with a little wiggle room is 0%).
November 20, 2025 at 11:14 AM
It matches anyone that is indexed to the listed accounts. The scale is a little random, only based on my matches (my worst cosine similarity with a little wiggle room is 0%).
I am a little hesitant on naming the embedder since not entirely sure about the licensing, but it is quick and fast with a slight quality trade off😅. I wanted to use Jina initially but pricing calculation made it infeasible.
November 20, 2025 at 11:10 AM
I am a little hesitant on naming the embedder since not entirely sure about the licensing, but it is quick and fast with a slight quality trade off😅. I wanted to use Jina initially but pricing calculation made it infeasible.
#PostgreSQL all the way down, with pgvector and pgvectorscale extensions. Embedding is done in rounds, collecting 25 only English posts&profile texts by a mixed score of recency and text length, then averaging for the account embedding.
November 20, 2025 at 11:06 AM
#PostgreSQL all the way down, with pgvector and pgvectorscale extensions. Embedding is done in rounds, collecting 25 only English posts&profile texts by a mixed score of recency and text length, then averaging for the account embedding.
It is embedding based, so you might change tone between accounts, probably post less diverse, so 89% is still pretty good. Also the percentage scale is based on my matches😇, so a little random. Thanks for trying it out!
November 20, 2025 at 10:38 AM
It is embedding based, so you might change tone between accounts, probably post less diverse, so 89% is still pretty good. Also the percentage scale is based on my matches😇, so a little random. Thanks for trying it out!
(It is just a demo, so will probably not be around for long in this form. But I thought it would provide value. E.g. “who did I not have on my radar”?)
November 20, 2025 at 9:43 AM
(It is just a demo, so will probably not be around for long in this form. But I thought it would provide value. E.g. “who did I not have on my radar”?)
Match scores are calculated by embedding cosine similarity, but in the detail view it gives you Euclidean distance and more stems, too.
November 20, 2025 at 9:41 AM
Match scores are calculated by embedding cosine similarity, but in the detail view it gives you Euclidean distance and more stems, too.
It is a tool comparing your match scores with attendees of the Eurosky conference. In your case it does not, since it respects your wish to not be shown to signed out users. But you can see my results here: divepool.social/c/eurosky/ra...
raedisch.net – Eurosky matcher
See which Eurosky participants match raedisch.net
divepool.social
November 20, 2025 at 9:39 AM
It is a tool comparing your match scores with attendees of the Eurosky conference. In your case it does not, since it respects your wish to not be shown to signed out users. But you can see my results here: divepool.social/c/eurosky/ra...
@ me or DM for anything, e.g. you do (not) want to be on the list. I will probably take it down in a couple of days or so.
November 20, 2025 at 9:13 AM
@ me or DM for anything, e.g. you do (not) want to be on the list. I will probably take it down in a couple of days or so.
Big thanks to @atproto.berlin, too, where I learned that I am doing backfilling almost right! (Except for the 3000 requests per 5 minute rate limit being per repo, not across all bsky repos, which is awesome! Thanks once more to #Bluesky.)
November 20, 2025 at 8:10 AM
Big thanks to @atproto.berlin, too, where I learned that I am doing backfilling almost right! (Except for the 3000 requests per 5 minute rate limit being per repo, not across all bsky repos, which is awesome! Thanks once more to #Bluesky.)
Or sports betting (gambling)?
November 20, 2025 at 6:27 AM
Or sports betting (gambling)?
Now with favicon and meta description divepool.social/c/eurosky/
EuroSky matcher – divepool
Compare your Bluesky account with EuroSky participants and find your closest matches
divepool.social
November 19, 2025 at 12:10 PM
Now with favicon and meta description divepool.social/c/eurosky/
That’s actually pretty cool. I have to train my politics brain, though.
November 19, 2025 at 10:42 AM
That’s actually pretty cool. I have to train my politics brain, though.
It is taking a mix of your latest and longest 25 English posts per embedding round, so each round it might improve. Rounds are a mechanism to ensure embedding breadth rather than depth. I only have one GPU running😅. Does the sorting go into the right direction for you?
November 19, 2025 at 10:40 AM
It is taking a mix of your latest and longest 25 English posts per embedding round, so each round it might improve. Rounds are a mechanism to ensure embedding breadth rather than depth. I only have one GPU running😅. Does the sorting go into the right direction for you?
it is embedding based and still indexing. So if you do not see any scores, you might not be in there yet. Or it might be a bug, or you only allow signed in viewers, or…
It also tries to extract some common keywords and more data if you click on the percentage.
It also tries to extract some common keywords and more data if you click on the percentage.
November 19, 2025 at 9:49 AM
it is embedding based and still indexing. So if you do not see any scores, you might not be in there yet. Or it might be a bug, or you only allow signed in viewers, or…
It also tries to extract some common keywords and more data if you click on the percentage.
It also tries to extract some common keywords and more data if you click on the percentage.
I can stop at any time. Just need Google and Stack Overflow. Oh, wait… .
November 19, 2025 at 6:53 AM
I can stop at any time. Just need Google and Stack Overflow. Oh, wait… .
It is doing all the coding.
November 19, 2025 at 6:49 AM
It is doing all the coding.