Building For You feed: https://bsky.app/profile/spacecowboy17.bsky.social/feed/for-you
Hobby project: linklonk.com
The results are:
-3.22% fewer "show less like this" (20,957 -> 20,282)
+17.24% more "show more like this" (4,083 -> 4,787)
+1.91% more likes (1,067,971 -> 1,088,331)
The results are:
-3.22% fewer "show less like this" (20,957 -> 20,282)
+17.24% more "show more like this" (4,083 -> 4,787)
+1.91% more likes (1,067,971 -> 1,088,331)
Would it be a new interaction type like "app.bsky.feed.defs#interactionDislike" or the existing "app.bsky.feed.defs#requestLess"?
@internet.bsky.social
We’re testing new systems to improve reply quality. See what’s coming: bsky.social/about/blog/1...
Would it be a new interaction type like "app.bsky.feed.defs#interactionDislike" or the existing "app.bsky.feed.defs#requestLess"?
@internet.bsky.social
I think to solve this we can let people opt out their account posts from appearing in For You.
If you want your account posts to not appear on For You feel free to DM me with "remove me from For You".
I think to solve this we can let people opt out their account posts from appearing in For You.
If you want your account posts to not appear on For You feel free to DM me with "remove me from For You".
If you ever need to go back to the recommendations you've seen in For You - use this simple feed:
If you ever need to go back to the recommendations you've seen in For You - use this simple feed:
With this change released I can clean up a bunch of hacks (based on scrolls and likes) I used to infer seen posts.
Everyone say thank you @danabra.mov
We've given the app a mini makeover with some new shades of blues! Plus, a whole bunch of bug fixes and performance improvements to make your experience better.
With this change released I can clean up a bunch of hacks (based on scrolls and likes) I used to infer seen posts.
I need to reboot the pc because Ubuntu updated nvidia drivers and I can no longer play games. That's one downside of serving the feed from your home pc.
I need to reboot the pc because Ubuntu updated nvidia drivers and I can no longer play games. That's one downside of serving the feed from your home pc.
Slightly fewer users were pressing "less like this": -1.4% (not statistically significant)
Slightly fewer users were pressing "less like this": -1.4% (not statistically significant)
rankthat.net/profile/spac...
This is still a client-side webapp talking directly to PDS. No AppView.
rankthat.net/profile/spac...
This is still a client-side webapp talking directly to PDS. No AppView.
All I needed to tell Claude was: "let's now do the heavy lifting and switch to real PDS storage"
Added my first item: pdsls.dev/at://did:plc...
All I needed to tell Claude was: "let's now do the heavy lifting and switch to real PDS storage"
Added my first item: pdsls.dev/at://did:plc...
Before committing to a lexicon I want to understand what a good user experience would look like and then build the corresponding lexicon
Before committing to a lexicon I want to understand what a good user experience would look like and then build the corresponding lexicon
Right now the replies are sorted by the number of likes. This is extremely gameable. 🧵
Right now the replies are sorted by the number of likes. This is extremely gameable. 🧵
I have no idea if this is a fair offer.
I have no idea if this is a fair offer.
The goal of this experiment is to reduce volatility of For You caused by your most recent likes.
I will share the results in a week or so.
The goal of this experiment is to reduce volatility of For You caused by your most recent likes.
I will share the results in a week or so.
The weight of your recent likes is spread out over fewer more recent likes. Detail are in the quoted post thread.
doesn't seem to remember very well that even if I've liked several of some type of post today, it should still keep showing me previous categories
The weight of your recent likes is spread out over fewer more recent likes. Detail are in the quoted post thread.
- 8.26% fewer "show less like this" (3340 -> 3064)
- 0.24% more posts in For You were liked (242438 -> 243024)
- 2.43% more feed loads (438867 -> 449537)
Per user and per request metrics:
- 8.26% fewer "show less like this" (3340 -> 3064)
- 0.24% more posts in For You were liked (242438 -> 243024)
- 2.43% more feed loads (438867 -> 449537)
Per user and per request metrics:
@gracekind.net
@gracekind.net
wscat --connect "wss://jetstream2.us-east.bsky.network/subscribe?wantedCollections=app.bsky.feed.post"
wscat --connect "wss://jetstream2.us-east.bsky.network/subscribe?wantedCollections=app.bsky.feed.post"
I don't know if this is just for me or everyone.
I don't know if this is just for me or everyone.
50% of the users on a given day will see popular posts demoted: score = score / power(popularity, 0.2)
50% will see no change (control arm).
"popularity" is the number of likes and reposts a post has.
50% of the users on a given day will see popular posts demoted: score = score / power(popularity, 0.2)
50% will see no change (control arm).
"popularity" is the number of likes and reposts a post has.
Curious? Then:
1. Go to linklonk.com/bluesky
2. Enter your user name
3. Enter the post url to explain
4. Hit Submit
Curious? Then:
1. Go to linklonk.com/bluesky
2. Enter your user name
3. Enter the post url to explain
4. Hit Submit
In the experiment the users are less likely to do all of these:
- show less like this
- show more like this
- like a post
Now that we have "show more like this" and "show less like this" we can tell not only which For You posts people like but also what they don't like. This is helpful to identify some potential algorithm improvements.
In the experiment the users are less likely to do all of these:
- show less like this
- show more like this
- like a post
@hailey.at would you know who could help with that?
Context below
@hailey.at would you know who could help with that?
Context below
The intention is to count likes and reposts of the same post only once per user.
From For You perspective we just care that user B endorsed item A.
For example, if you like and repost something it should be treated the same if you just like or just repost.
The intention is to count likes and reposts of the same post only once per user.
From For You perspective we just care that user B endorsed item A.
For example, if you like and repost something it should be treated the same if you just like or just repost.