Daily Summary (9pm UTC)

Today on meta.discourse.org

Key Stats

  • New posts (last 24h): 80
  • New topics (last 24h): 9

Top users (posts / likes in the last 24h)

User Posts Likes
Moin 7 7
Falco 5 9
Lilly 5 5
chapoi 4 8
martin 2 3
david 2 3
Jonathan5 1 4
awesomerobot 1 4
PandaKai7 1 4
system 1 3

Busiest conversation clusters (with sources):

  • Policy + privacy tradeoffs around age checks in Support: read more
  • UX debate over infinite scroll vs pagination (and what “bottomlessness” really means) in Support: read more
  • Multiple bug fixes/triage, including duplicate mention “pings” (notably in chat): read more
  • Frontend build/deprecation fallout affecting themes/plugins (now fixed and backported): read more
  • Hosting/admin troubleshooting: YJIT enablement clarity in Support: read more

Also spotted in the stream (quick hits, with links):

  • Plugin maintenance: the Contact Form plugin received an update to restore compatibility with newer Discourse versions: read more
  • Admin docs nuance on “permanent delete” safeguards being potentially underspecified: read more
  • A long-running plugin thread resurfaced with a request to disable the animation in Discourse Who’s Online: read more
  • “Read-only mode” UI questions popped up again (and may differ by UI/state): read more
  • Form template UX request: allowing copy/paste inside template-controlled fields: read more
  • Feature brainstorming on “staff approval” workflows for wiki edits continued: read more
  • A feature bump request for embedding chat elsewhere (e.g., as an entrypoint for a radio community): read more
  • Self-hosting auth integration friction: OIDC users not linking to existing accounts by email: read more
  • Docker / TLS onboarding: a fresh install thread was marked solved (Let’s Encrypt template): read more

Interesting Topics

Support

  • RGJ’s age requirement thread broadened into a practical/ethical discussion about verification services, privacy risk, and what sites can reasonably enforce in policies; awesomerobot highlighted the threat model of third-party data storage and the appeal of zero-knowledge approaches: read more (related theme component: read more)

  • A newly migrated community asked how to avoid infinite scroll on the homepage topic list to keep a footer accessible; chapoi confirmed “pagination isn’t on the roadmap” and suggested custom landing pages instead, while others proposed changing the homepage mode and discussed whether infinite scroll is inherently harmful: read more (footer component mentioned: read more)

  • A Windows 7 support follow-up explored what users see on unsupported browsers and whether admins can customize that message; the thread narrowed into the subtlety that strings under js.* may not load when JS is intentionally blocked, limiting customization: read more (background/related: read more, text customization guide: read more)

  • An [Admin Notice] about id:discourse.user.userOptions led to a hunt for which theme/component/plugin triggered it; the OP correlated it with notification oddities and started isolating customizations and grepping plugin codepaths: read more (broader plugin bundling context: read more)

  • AI translation behavior and cost controls came up again: a site hit token/day rate limits while translations kept retrying, and the OP asked whether “thinking/vision/reasoning effort” affects translation outcomes in Discourse AI: read more

#Self-hosting

  • A “YJIT enablement doesn’t work” report turned into a precise debugging lesson: environment variables can be scrubbed by sudo (unless -E is used), which made YJIT appear disabled; the thread ended with a confirmed fix and the solution marked: read more

  • A backup restore failed with a Sidekiq renice permission error and a Postgres type collision (ai_moderation_setting_type already exists), prompting advice to ensure the target DB is truly clean and to adjust the restore workflow: read more (referenced prior incident: read more)

bug

  • A chat mention bug produced duplicate “pings” pointing to the same message; staff confirmed reproduction and shipped a fix shortly after: read more

  • Deprecation handling broke builds for some customizations: missing ember-this-fallback/* modules caused runtime errors and made it harder to locate the real offending code; the issue was investigated, fixed, and queued for backport: read more (example fallout in a theme thread: read more)

  • A new AI docked composer experience was reported as visually polished but functionally unstable: editing controls blocked, markdown/preview UX oddities, quoting breaking sidebars, and mobile scrolling/lockups—reproduced even in safe-mode with only official plugins: read more

  • The long-running “Proofread breaks quotes” issue got reaffirmed as fixed, with a pointer to the upstream change and follow-up testing around models/prompts—particularly important for communities where quotes must remain verbatim: read more

#Announcements

  • Trust system mechanics were clarified around the “upcoming change” for trust level 3 defaults: overrides remain preserved, but using Reset reveals the new defaults, and disabling the upcoming change restores older defaults (until the transition is finalized): read more

  • Discourse’s full-site embedding as a comment system was explicitly framed as a Disqus replacement (with data/control benefits), including a callout to member-only communities that want public blog engagement without anonymous posting: read more

#customization:extras

  • A product idea pitched dead Amazon/eBay link detection and replacement with related products; staff pushed for real-world Discourse testing results before it can be evaluated as broadly useful: read more

blog

  • The companion discussion thread for Discourse Discover: Hobbyist Edition landed, showcasing niche communities (unicycling, tiny houses, flight sim) and the kind of “specific expertise” that Discourse instances tend to accumulate: read more

Activity by the @team Group

  • awesomerobot weighed in on the risks of third-party age verification (especially services that store sensitive identity data) and why zero-knowledge verification could reduce exposure—while still acknowledging transfer risk: read more

  • chapoi answered a roadmap question bluntly (“no pagination plan” for topic lists), suggested workarounds like custom landing pages, and later pushed back on equating Discourse’s infinite scroll with addictive algorithmic feeds: read more, read more

  • lindsey confirmed the duplicate mention/ping bug affected staff too and kicked off investigation: read more

  • Falco closed the loop on the ping issue with a shipped fix, clarified YJIT toggling behavior (and what the env var does/doesn’t do), explained why “unsupported browser” messaging can’t rely on JS-loaded strings, reaffirmed the fix for AI-proofread quote preservation, and positioned embedded comments as a Disqus alternative:

  • david investigated the ember build/deprecation regression, confirmed expectations (“deprecated but still supposed to work”), and delivered a fix with a backport plan: read more, read more

  • martin clarified how Discourse’s site-setting override system interacts with the new TL3 defaults (especially what Reset does and how “upcoming changes” affect which default set you see): read more

  • Bas challenged an Extras pitch constructively by asking for real Discourse instance testing and results before further evaluation: read more


If yesterday proved anything, it’s that Meta can debate bottomlessness while still shipping fixes—may all your pings be single, your YJIT be truly enabled, and your scroll be responsibly unaddictive.