Version packages (#4336) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

explicit-slurs: remove stray match for unrelated word (#4330)

bsky: improve error messaging for DPoP bound access tokens (#4333)

Remove non-existing `reporter_stats` from materliaziled view to refresh (#4335)

As blue confetti settles in Los Angeles after an historic World Series win, we close the chapter on another electric sports event on Bluesky. It’s during these cultural flashpoints when Bluesky is at its best – when stadium crowds are going wild, you can feel that same excitement flooding into posts. Numbers can help describe the scale of that feeling. There were approximately 600,000 baseball posts made during the World Series, based on certain key terms. (note: we’re pretty sure that this number is an undercount, as it’s hard for us to accurately attribute to baseball the flood of “oh shit” posts that came in during Game 7) At least 3% of all posts made on November 1 (Game 7) were about baseball. The final game also resulted in a +30% bump in traffic to Bluesky, with engagement spikes up to +66% from previous days. We loved seeing World Series weekend in action, but it wasn’t a total surprise. In the last three months, sports generated the third-highest number of unique posts of any topic. Sports posters are also seeing greater engagement rates from posting on Bluesky than on legacy social media apps - up to ten times better . But in a world of analytics, it’s easy to lose the love of the game. In that regard, we’re fortunate to have a roster of posters who bring the intangibles. They genuinely care about sports. Less hate, more substance and celebration. yep, this is the baseball place now [image or embed] — Keith Law ( @keithlaw.bsky.social ) November 1, 2025 at 10:28 PM That was the greatest baseball game I’ve ever seen. — Molly Knight ( @mollyknight.bsky.social ) November 1, 2025 at 9:19 PM If this World Series proved anything, it’s that big moments are more enjoyable when they unfold in real time, with real people. Sports has the juice on Bluesky — and every high-stakes game is bringing more fans into the conversation.

[APP-1484] Add thread/search exploration code and feature gate (#4327) * Add ranking exploration code and feature gate * Import sort

resolveLexicon endpoint (#4227) * Add resolveLexicon.json for lexicon resolution * Update lexicons/com/atproto/lexicon/resolveLexicon.json * chore: formatting and linting * chore: Adding changeset information * documentation: Updating descriptions for properties. * refactor: Updating com.atproto.lexicon.resolveLexicon lexicon schema from PR discussion and feedback * Update lexicons/com/atproto/lexicon/resolveLexicon.json Co-authored-by: Nick Gerakines <[email protected]> * tweak changeset * lexicon codegen --------- Co-authored-by: Nick Gerakines <[email protected]> Co-authored-by: Matthieu Sieben <[email protected]>

At Bluesky, we’re building a place where people can have better conversations, not just louder ones. We’re not driven by engagement-at-all-costs metrics or ad incentives, so we’re free to do what’s good for people. One of the biggest parts of that is the replies section. We want fun, genuine, and respectful exchanges that build friendships, and we’re taking steps to make that happen. So far, we’ve introduced several tools that give people more control over how they interact on Bluesky. The followers-only reply setting helps posters keep discussions focused on trusted connections, mod lists make it easier to share moderation preferences, and the option to detach quote posts gives people a way to limit unwanted attention or dogpiling. These features have laid the groundwork for what we’re focused on now: improving the quality of replies and making conversations feel more personal, constructive, and in your control. In our recent post , we shared some of the new ideas we were starting to develop to encourage healthier interactions. Since then, we’ve started rolling out updates, testing new ranking models, and studying how small product decisions can change the tone of conversations across the network. We’re testing a mix of ranking updates, design changes, and new feedback tools — all aimed at improving the quality of conversation and giving people more control over their experience. Social proximity We’re developing a system that maps the “social neighborhoods” that naturally form on Bluesky — the people you already interact with or would likely enjoy knowing. By prioritizing replies from people closer to your neighborhood, we can make conversations feel more relevant, familiar, and less prone to misunderstandings. Dislikes beta Soon, we’ll start testing a “dislike” option as a new feedback signal to improve personalization in Discover and other feeds. Dislikes help the system understand what kinds of posts you’d prefer to see less of. They may also lightly inform reply ranking, reducing the visibility of low-quality replies. Dislikes are private and the signal isn’t global — it mainly affects your own experience and, to an extent, others in your social neighborhood. Improved toxicity detection Our latest model aims to do a better job of detecting replies that are toxic, spammy, off-topic, or posted in bad faith. Posts that cross the line are down-ranked in reply threads, search results, and notifications, reducing their visibility while keeping conversations open for good-faith discussion. Reply context We’re testing a small change to how the “Reply” button works on top-level posts: instead of jumping straight into the composer, it now takes you to the full thread first. We think this will encourage people to read before replying — a simple way to reduce context collapse and redundant replies. Reply settings refresh Bluesky’s reply settings give posters fine-grained control over who can reply, but many people don’t realize they exist. We’re rolling out a clearer design and a one-time nudge in the post composer to make them easier to find and use. Better visibility means more people can shape their own conversations and prevent unwanted replies before they happen. Conversations you start should belong to you. We won’t get everything right on the first try. Building healthier social media will take ongoing experimentation supported by your feedback. This work matters because it tackles a root flaw in how social platforms have been built in the past — systems that optimize for attention and outrage instead of genuine conversation. Improving replies cuts to the heart of that problem. Over the next few months, we’ll keep refining these systems and measuring their impact on how people experience Bluesky. Some experiments will stick, others will evolve, and we’ll share what we learn along the way.

fix wildcard endings in firehose filtercollections (#4038) * allow filtercollections to have a wildcard ending * changeset * fix parameters

Add optional via property to follow records (#4317)

Search filtering by tag (#4312)

oauth example app rework (#4319) * oauth example app rework * changeset

Version packages (#4307) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Version packages (#4302) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Set dark background on authorization pages `<body>` in dark mode (#4301) * Set dark background on authorization pages `<body>` in dark mode * tidy

Add missed changeset (#4314)

[TS-160] Add `debug` field to `PostView` and `ProfileView*`s (#4297) * Add debug field to lex and proto, codegen * Include debug field for DIDs configured via env var * Gen bufs * Update types and view * Add debug field to profile views * Safely populate debug field * Format * Use util * Simplify existence check * Values may be undefined * Clarify type * Update descriptions * Include default value for dev-env * Add test harness (cherry picked from commit 7e49b4cf4cf3cbb989114540081e74f7a34acb6f) * Use unknown field type instead of serializing * KISS keep it set stupid * Fix comments * Differentiate between source of debug * Add matching profile view test * Specifying types of the merged objects isn't really helpful * Changeset * Add comment * Put includeDebugField on HydrateCtx and use in views * Mutate TestNetwork innards for tests * Format * Use separate debug schema for profile test * Little more terse * Oops, remove last usage * Remove vestigial props * Include other actor metadata * Simply views, remove debugField

Fix: Label creation notification race condition (#4304) Previously the notification for the label creation was dispatched before the label was actually inserted into the database. In all other cases where we use postgresql's notify, we do it after the insertInto, such that the listener is guaranteed to receive the newly created record when querying

ozone: check labels when asserting protected tags (#4283) * check labels when asserting protected tags * fix * fix' --------- Co-authored-by: Devin Ivy <[email protected]>

Version packages (#4320) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Clean up `getPostThreadV2` in prep for future work (#4269) * Deprecate prioritizeFollowedUsers, codegen * Remove from agent response types * Remove codepaths * Format * Changeset * Remove fields entirely * Changeset * Swap changeset

resolveLexicon endpoint (#4227) * Add resolveLexicon.json for lexicon resolution * Update lexicons/com/atproto/lexicon/resolveLexicon.json * chore: formatting and linting * chore: Adding changeset information * documentation: Updating descriptions for properties. * refactor: Updating com.atproto.lexicon.resolveLexicon lexicon schema from PR discussion and feedback * Update lexicons/com/atproto/lexicon/resolveLexicon.json Co-authored-by: Nick Gerakines <[email protected]> * tweak changeset * lexicon codegen --------- Co-authored-by: Nick Gerakines <[email protected]> Co-authored-by: Matthieu Sieben <[email protected]>

Search filtering by tag (#4312)

At Bluesky, we’re building a place where people can have better conversations, not just louder ones. We’re not driven by engagement-at-all-costs metrics or ad incentives, so we’re free to do what’s good for people. One of the biggest parts of that is the replies section. We want fun, genuine, and respectful exchanges that build friendships, and we’re taking steps to make that happen. So far, we’ve introduced several tools that give people more control over how they interact on Bluesky. The followers-only reply setting helps posters keep discussions focused on trusted connections, mod lists make it easier to share moderation preferences, and the option to detach quote posts gives people a way to limit unwanted attention or dogpiling. These features have laid the groundwork for what we’re focused on now: improving the quality of replies and making conversations feel more personal, constructive, and in your control. In our recent post , we shared some of the new ideas we were starting to develop to encourage healthier interactions. Since then, we’ve started rolling out updates, testing new ranking models, and studying how small product decisions can change the tone of conversations across the network. We’re testing a mix of ranking updates, design changes, and new feedback tools — all aimed at improving the quality of conversation and giving people more control over their experience. Social proximity We’re developing a system that maps the “social neighborhoods” that naturally form on Bluesky — the people you already interact with or would likely enjoy knowing. By prioritizing replies from people closer to your neighborhood, we can make conversations feel more relevant, familiar, and less prone to misunderstandings. Dislikes beta Soon, we’ll start testing a “dislike” option as a new feedback signal to improve personalization in Discover and other feeds. Dislikes help the system understand what kinds of posts you’d prefer to see less of. They may also lightly inform reply ranking, reducing the visibility of low-quality replies. The signal isn’t global — it mainly affects your own experience and, to an extent, others in your social neighborhood. Improved toxicity detection Our latest model aims to do a better job of detecting replies that are toxic, spammy, off-topic, or posted in bad faith. Posts that cross the line are down-ranked in reply threads, search results, and notifications, reducing their visibility while keeping conversations open for good-faith discussion. Reply context We’re testing a small change to how the “Reply” button works on top-level posts: instead of jumping straight into the composer, it now takes you to the full thread first. We think this will encourage people to read before replying — a simple way to reduce context collapse and redundant replies. Reply settings refresh Bluesky’s reply settings give posters fine-grained control over who can reply, but many people don’t realize they exist. We’re rolling out a clearer design and a one-time nudge in the post composer to make them easier to find and use. Better visibility means more people can shape their own conversations and prevent unwanted replies before they happen. Conversations you start should belong to you. We won’t get everything right on the first try. Building healthier social media will take ongoing experimentation supported by your feedback. This work matters because it tackles a root flaw in how social platforms have been built in the past — systems that optimize for attention and outrage instead of genuine conversation. Improving replies cuts to the heart of that problem. Over the next few months, we’ll keep refining these systems and measuring their impact on how people experience Bluesky. Some experiments will stick, others will evolve, and we’ll share what we learn along the way.

fix wildcard endings in firehose filtercollections (#4038) * allow filtercollections to have a wildcard ending * changeset * fix parameters