Today on meta.discourse.org…
Key Stats
Over the past 24 hours, the community produced 51 new posts across 27 topics, earning 48 likes from fellow members. The most-liked discussion was How can I see AI-translated posts? with 7 likes, while the ever-popular Free plan beta thread kept churning with questions about LLMs, trust levels, and closed communities. #Plugin and ux tied as the busiest categories, each hosting five distinct topics in the last day.
Top contributors by posts and likes:
| User | Posts | Likes |
|---|---|---|
| Moin | 5 | 9 |
| NateDhaliwal | 3 | 6 |
| Jagster | 5 | 4 |
| Darkpixlz | 4 | 3 |
| Satonotdead | 4 | 3 |
| Falco | 2 | 3 |
| Don | 1 | 3 |
| Jimmy0017 | 1 | 3 |
| Soul | 1 | 3 |
| Dowlass | 4 | 2 |
Interesting Topics
#Announcements
-
Dowlass confessed in the Discourse Free plan beta thread that “LLM…my acronym for the day!” was a brand-new discovery, and admitted that trust-level jargon feels like a foreign language for a small private community. NateDhaliwal pointed to the detailed Trust Levels guide, and Moin offered a patient, jargon-free explanation of groups versus trust levels for non-technical admins.
-
One1 discovered chat notifications can have sound packs and asked whether a preview button could be added to settings — only to learn from Falco that selecting a pack already plays a preview right in the preferences page.
-
Darkpixlz confirmed that the composer toolbar overlap they had flagged earlier in the Composer Redesign testing thread has been fixed, applauding the team with a well-earned clap emoji: “seems like this was fixed, nice work.”
Support
-
Jagster was puzzled that AI-translated posts stopped appearing in English after switching the UI locale, and Moin quickly diagnosed the culprit: the Content languages preference suppresses translations for any language you claim to understand. Removing Finnish from the list instantly brought the content-localization translations back, though Jagster noted the behavior had silently changed compared to earlier builds.
-
Pfaffman shared a helpful tip for the admin notice about outdated theme/plugin code, noting there is a setting to auto-update the code on upgrade rather than dealing with the warning manually.
-
Satonotdead opened a discussion about moving category boxes to the top, arguing that subcategory boxes interfere with the daily UX experience and that the core behavior could be broadly rethought — revisiting an earlier category-boxes design discussion and lamenting that CSS tricks alone can’t fix it.
#Plugin
-
Jimmy0017 reported that regular users viewing another member’s profile could hit a scary permission error when the See TL3 Progress plugin is installed — then went ahead and opened a pull request to fix it. NateDhaliwal appreciated the catch and merged the PR promptly.
-
Soul launched Discourse-indexnow, a new plugin that instantly notifies Bing, Yandex, and other IndexNow-compatible search engines when topics are created or edited. It ships with localized URL support, historical backfill, hourly/daily submission caps, automatic key rotation, and a detailed admin log panel showing success/failure trends.
-
NateDhaliwal asked whether Resenha’s 1-to-1 video calls can be restricted to specific groups to prevent spam avenues, and Falco confirmed the feature is disabled by default and can be gated behind chosen groups.
-
Akanora hit a fatal error with the Animated Avatar plugin on v2026.8.0-latest.1: a missing
resize_animatedkey crashes the entireUploadCreatorpipeline, breaking all avatar uploads — not just GIFs. A dedicated bug report was opened with a full Sidekiq backtrace to help the plugin author reproduce it.
ux
-
Satonotdead suggested simplifying the subcategory pills, arguing “Community Building looks good! But Community Building > Praise not so much” and that AI-era users want simple, direct links. Orioni pushed back with a real-world edge case: product categories using shared subcategory names like “installation” or “support” would lose essential context. Mcwumbly agreed the current rendering should be made nicer and linked a related feature request about more concise inline subcategory links.
-
Darkpixlz flagged a confusing notification when a topic you replied to gets solved: the notice looks identical whether you provided the solution or someone else did. Their proposal is to differentiate strings like “Your reply solved this topic!” versus “Topic solved!” — or offer a way to disable the notice entirely. Moin pointed to the existing notification preference for solved topics, though Darkpixlz called it a “nuclear option” that also kills the thank-you notification, linking back to the earlier feedback thread on this feature.
-
Philh untangled the head-spinning double negative in the 2 setting description, proposing clearer wording: “Manage S3 backup retention externally — when enabled, Discourse will not automatically delete old S3 backups.” A small change that saves everyone a headache.
#Site Management
- In the long-running guide to configuring a closed or private community, Dowlass asked how “invite-only” actually works in practice — do staff walk up to friends with a signup link, or are invites sent by email? Moin explained the invite modal and sidebar flows, noting admins get extra options and that invite links can be shared anywhere, not just via email.
#Self-hosting
- Joshua8600 reported a regression after upgrading from February 2026 to the July 2026 release: auto-bumped topics in their Support category now move to the top of the list without any unread indicator or bolded title, so moderators miss them entirely. Jagster offered a thought-provoking response — isn’t the request, in effect, that auto-bumps should show a “false unread status” to everyone?
Activity by the Team Group
-
Moin was the day’s support superhero, untangling the AI translation visibility mystery, explaining notification preferences for solved topics, walking Dowlass through invite-only community setup, and offering a patient, jargon-free intro to trust levels in the Free plan beta thread.
-
Falco fielded feature questions with quick, definitive answers, confirming that chat notification sound previews already play on selection and reassuring NateDhaliwal that Resenha’s 1-to-1 calls are restricted by group and disabled by default.
-
Nat followed up on user reports in the Conversational AI Search thread, asking for settings confirmation to diagnose why the feature “does not work” for some members.
-
Mcwumbly joined the subcategory-pills UX conversation, agreeing the styling should be made more concise, and rounded out the day by sharing a NYT interview with Priya Parker on the art of conflict in the Community Building forum.
From AI translations hiding in language preferences to auto-bumps tiptoeing silently to the top of topic lists, yesterday’s biggest mysteries just needed the right settings toggle — and maybe a well-aimed pull request! 
Closed label in group cards, arguing it reads like “archived”, especially in Chinese; 
