kairo162 noticed the label “regular” showing under usernames and across profiles; Lilly connected it to trust-level naming (post), and Moin clarified it’s a selectable Title (often from trust-level badges) that users can enable in preferences (solution). read more
In Chinese-language support, jack222 reported that anonymous visitors can’t view /activity/reactions and suggested either enabling read-only access or improving the error message (report); Moin pointed to an existing UX request discussing exactly that permission/UX gap (reply) and referenced the related feature request (Allow /activity/reactions to be visible by anon). read more
cshll hit a “Badges” page failure and suspected configuration issues (report); NateDhaliwal suggested checking the enable_badges site setting (reply), while Noble_Fish asked whether it duplicates an earlier thread (dup-check) and probed whether the install looked non-standard (follow-up). read more
cshll also posted an SEO/validation screenshot about “inspecting Live URL,” but without details yet—likely a “needs repro steps” thread unless more context lands (topic). read more
Noble_Fish requested global defaults for “New topic approval” and “New reply approval,” noting that current global settings only cover Everyone except selected groups and don’t match the workflow where a subset of users later gets moved into “must be reviewed” mode across all categories (feature request). read more
In a self-hosting migration discussion, Lilly shared a recent real-world migration (Ubuntu 24.04 → Ubuntu 24.04) and highlighted practical tips like shortening DNS TTL and planning for backup upload time (migration notes). read more
janbolmeson reported that @discourse/mcp fails when connecting to Discourse installed in a subfolder (e.g. /forum) because tool calls drop the subpath—breaking validation and subsequent API calls (bug report). read more
In an ongoing structured-data issue, Richie noted Google tests still failing for three URLs due to missing mainEntity.suggestedAnswer.text, despite partial improvements (status update). read more
AI translation troubleshooting continued: RBoy encountered a 400 error tied to unsupported json_schema response formatting with a model change (error details), and Falco explained translations need a model that supports structured outputs (common in recent SOTA models) (guidance). read more
#Theme component
Quote Callouts got a UX + edge-case sweep: nathank observed rendering issues in excerpts/Oneboxes (note), scavin flagged the default shortcut as risky on macOS because users may hit ⌘+Q and quit the browser (macOS concern), and Arkshine weighed alternative key combos while tracking excerpt/Onebox behavior (response). read more
#Announcements
The “Upcoming Changes” notification system drew feedback on notification noise: Lilly asked if alerts can be disabled without CSS (preferring a setting-level solution) and shared a temporary CSS workaround (request + snippet); NateDhaliwal pointed out a likely remaining unread-count indicator even if notifications are hidden (reply). read more
#Plugin
The Locations plugin thread got a substantial update: merefield announced a fix now deployed (plus compatibility work) and described how Early Access differs from the public version—including a 3D globe users map and IP-based location lookup (release notes); gilles thanked them for the quick turnaround (thanks). read more
Theme authoring guidance discussions continued: BrianC asked how to approach forking a theme while ensuring the parent stays updated over time—raising the common long-term maintenance question for serious customizations (question). read more
A complex “facts from logs” question landed: Sukenik asked how a forensic expert might establish whether historical suspensions carried over during a 2023 migration from Discourse to a custom forum (scenario); Andrew_Rowe highlighted how hosting model and log retention matter, and that three years is a long time for server logs to survive (initial response). read more
Falco helped unblock AI translation troubleshooting by explaining that translation workflows require a model that supports structured outputs, and that most modern SOTA models now include this capability (AI translations guidance, context in error report).
If you made it through today’s feed without accidentally hitting ⌘+Q while trying to insert a callout, you’ve earned your Regular title the hard way.
New posts (last 24h): 97 (spread across threads like the iOS PWA push investigation in read more and the new bulk tag workflow announcement in read more)
New topics (last 24h): 13 (including new reports like the Edge mobile modal sizing issue in read more, plus new how-tos like standardizing sidebar categories in read more)
Publishing a sidebar-standardization how-to in read more
Interesting Topics
#Announcements
Staff introduced a cleaner bulk-tag workflow: a single Manage Tags modal replaces separate add/replace/remove actions, prompting follow-up feedback about revision history clarity and missing UI affordances in Better bulk tag management. Related: the missing tag info button discussion in read more and the older feature request this essentially fulfills in read more.
#Site Management
A practical admin recipe for making everyone’s sidebar categories consistent (by resetting default navigation menu categories and applying “historically”) landed in How to set the same sidebar categories for all the users, followed by questions about why the “two-step + wait” approach matters in read more.
A detailed report found iOS pwapush-notifications subscription can silently fail when the service worker isn’t controlling the app yet, complete with repro steps and a proposed fix PR in Push notification subscription in iOS PWA silently failing…. There’s active discussion on whether adding skipWaiting() / clients.claim() to the main SW is the right long-term move in read more.
A gifs integration snag: when “Limit infinite search results” is enabled, Klipy categories trigger loadMore incorrectly and can error due to missing query params, with a plugin PR suggested in Klipy issue when “Limit infinite search results” is set. This also cross-connects with earlier Klipy oddities like repeated results in read more.
A localization/SEO fix is underway: a PR was filed to ensure hreflang stays compliant while preserving internal locale params in Unerwarteter Sprachcode.
Mobile Edge users spotted a cramped “Permanent delete” confirmation dialog; staff confirmed it’s an older modal codepath that hasn’t moved to the modern d-modal yet in 手机微软edge浏览器“永久删除”弹出窗口太窄, with confirmation in read more.
A support thread dug into why a automation job (Data Explorer PM scheduling) triggers manually but won’t fire on schedule—debugging ideas ranged from time zones and “missed jobs” to logs and possibly special characters in names in Debug Why Timed Automation is not Automatically Running. The “monthly vs frequent interval” suspicion sharpened later in read more. (Background reference: Discourse Automation plugin in read more.)
Admins wanting to remove “Log in with your access key” were pointed toward the hidden enable_passkeys setting and/or CSS approaches in How to remove “Log in with your access key”, with hidden settings explained in read more.
#Self-hosting
A migration/restore failure involving Cloudflare R2 and checksum incompatibilities surfaced during a server move; troubleshooting and environment details are collected in 迁移Discourse恢复数据失败, including an R2-specific env-var suggestion in read more.
A “badges error launching” support thread turned into a broader self-hosting health-check: plugin overload, incorrect SMTP settings, and risky storage choices were called out in Badges Error launching, with concrete “remove unofficial plugins first” guidance in read more.
chapoi confirmed the Edge mobile delete-modal is legacy UI and praised the report in read more, continued triage on the repeated-upload-in-RTE report in read more, and helped narrow self-hosting/plugin troubleshooting steps (including safe mode) in read more. chapoi also pulled nat into localization/quote metadata concerns in read more, and confirmed a fix got merged for reaction long-press behavior in read more.
fzngagan posted a clear step-by-step guide for enforcing consistent sidebar categories for all users in read more, sparking follow-up testing/edge-case questions from others in read more.
lindsey announced the consolidated bulk tag management modal and invited feedback in read more, which immediately surfaced UX expectations around tag tooling and revisions in read more.
nat analyzed why translated quotes lose useful source info (with German being hit hardest) and noted more evaluation work is needed in read more.
sam linked a fix PR for a reported MCP subfolder-install issue in read more.
manuel clarified how the “skills” track and the Blocks API should remain maintainable over time (and what to expect if you fork examples) in read more.
awesomerobot acknowledged the “Permanent” filter confusion on Upcoming Changes and explained it as an optional workflow step that was skipped for some items in read more.
gerhard proposed a concrete Cloudflare R2 workaround using AWS checksum-related env vars (plus rebuild steps) for restore failures in read more.
HAWK responded to an SEO console “Live URL inspection” concern, suggesting the issue is likely due to new/empty sitemaps and minimal public content in read more.
supermathie requested environment details (object store specifics) to unblock troubleshooting for the restore/migration failure in read more.
Yesterday’s theme was “tiny modals, missing automations, and service workers refusing to take charge”—may today’s builds be smooth, your tags tidy, and your push subscriptions actually land where they should.
In Introducing Upcoming Changes, Lilly shared a way to reduce “admin noise” by dismissing notifications (and confirmed behavior changes discussed in-thread) — a practical follow-up for anyone piloting feature rollouts via upcoming-changes.
A tricky embedding case: Pavel_Kravets tried to embed Discourse into Salesforce Experience Cloud via iframe/full-app mode; Falco outlined the practical paths forward (shared base domain or adjusting same_site_cookies) in Salesforce Community iframe embedding.
A REST API regression report: /admin/users/list/active.json intermittently 500s due to main_user_record[:silenced_till] nil; discussion centered on upgrading and a prior fix landing later than v3.5.2 in API calls fail due issue with main_user_record silenced_till (with zogstrip pointing to the likely fix here).
Redirect UX: admins asked how to redirect visitors who hit non-existent pages; the community pointed to a canonical solution thread in How to make a redirect from non-existent pages, referencing the “redirect all 404 pages” approach (read more).
A reproducible editor/navigation bug (reported in Chinese): after expanding replies → “jump to post” → edit → save, users can hit “URL/resource not found”; chapoi confirmed and zogstrip started a fix in 出错了：找不到请求的 url 或资源 (related duplicate report: 出错了：找不到请求的 url 或资源。).
Bot report fatigue: the “Web Crawler User Agents” report has ballooned, driven by Qwantbot variants whose numeric suffix changes; suggestion was to consolidate these strings in reporting in Web Crawler user agents - Qwantbot agents growing in number.
A product-direction check-in: a third-party asked whether communities want gamification beyond trust levels (XP, streaks, currencies, progression paths) and whether an integration is worth building in Gamification for Discourse communities beyond trust levels.
chapoi jumped on two fronts: confirmed and narrowed down a reproducible navigation/edit bug in 出错了：找不到请求的 url 或资源, and reinforced trust-level integrity (no “reading time scripts”) in 怎么刷阅读时长？ (and follow-up clarification in post 6).
supermathie supported a migration troubleshooting thread by pointing to the relevant guide and asking for a redacted ENV block to diagnose further in 迁移Discourse恢复数据失败.
Yesterday’s theme was making things smaller, smoother, and less surprising—may your GIFs shrink into WebPs, your automations land on the right day of the month, and your 404s politely find their way home.
A lot of the conversation clustered around major product updates like nested replies (read more) and improvements to solved workflows via the new “Me too” button (read more), plus a steady stream of bug reports and fixes (e.g., the grid lightbox order issue: read more).
Introducing nested replies — markvanlan unveiled an early look at Discourse’s new fully-threaded topic view and the admin controls behind it (read more). Follow-up discussion covered practical rollout concerns like bulk-toggling existing topics (read more), reversibility and how nested mode is represented in the DB (read more), and why threading is intentionally not a per-user preference (read more).
Solved improvements: “Me too” for unsolved support topics — lindsey announced a new “Me too” button for solved topics in support categories to help members signal impact and opt into solution notifications (read more). Feedback quickly turned toward admin-side reporting/threshold alerts (read more) and whether “Me too” should bump older threads (read more); awesomerobot also clarified how it differs from experimental “boost” behavior (read more).
Granular group-based permissions for anonymous and logged in users — martin explained the deprecation of the confusing @everyone pseudogroup in favor of two explicit automatic groups: anonymous (ID 4) and logged_in_users (ID 5) (behind an upcoming change) (read more). A key clarification: this won’t unexpectedly lock down public categories, because category/tag permissions treat “everyone” differently (read more).
Grid lightbox images show in the wrong order (lightbox) — Shauny reported that arrow-key navigation in a grid-based lightbox didn’t follow upload order (read more). zogstrip traced it, fixed it, and linked the core PR for review (read more).
Users cannot see who is a group owner — A regression left “group owner” info present in page data but no longer rendered for regular users (read more). awesomerobot confirmed it was likely an unintentional change from a 2023 refactor and opened a PR to restore visibility (read more).
Norwegian locale mismatch in AI locale detection vs content-localization locales (ai) — A report showed Norwegian being detected as no while translation targets used nb_NO/nb-NO, causing near-identical translations and wasted credits (read more). nat suggested a quick mitigation by duplicating the locale detection agent and adjusting its prompt (read more); the reporter found nb-NO wording worked best in practice (read more).
Assign/Re-assign on mobile: can’t type in the search field (mobile, assign) — Continued reports suggest the assign modal can error out on mobile, preventing typing and producing JS exceptions (including attempts at safe-guarding targetElement) (read more).
Incorrect ActivityPub handler addresses shown in /ap/about (activitypub) — A small but impactful formatting bug: displayed handles were missing a leading @, making them harder to copy into ActivityPub clients (read more).
Discourse-patch-triage bot: how to work with forks — Lilly asked how to collaborate effectively with the new discourse-patch-triage bot, especially when the PR is from a fork and the bot suggests changes (read more). isaac clarified how to invoke it (@patch-triage + plain English), what it can do, and that it cannot commit onto forks (and promised clearer messaging + internal discussion about permissions) (read more, read more).
Potential upcoming plugin name clash: “Workflow” — merefield flagged a possible collision: a core PR appears to introduce “Workflow” while a long-standing plugin with the same name already exists, risking admin UI confusion (read more). Related background: the existing plugin’s Meta topic and ongoing compatibility updates (read more).
MiniScheduler + Sidekiq 8 integration PR merged and released — A quick turnaround: dmke requested another look, and david confirmed the PR was merged and released (read more, read more).
Hide user cards for anonymous users (solved) — In a French-language support thread, gilles asked how to prevent user cards from displaying to anonymous visitors (read more). chapoi pointed to the relevant site setting and the topic was marked solved (read more).
Translation only translates the title, not the post (content-localization) — A report described post body translation failing while topic titles translated normally (read more). nat noted a post-length limit can block translation and indicated it could likely be increased; shortly after, the content did translate (read more).
Fresh install: activation email not sent (but test email works) (email) — A self-host admin found that registration/activation emails didn’t arrive even though discourse-doctor test email delivered immediately (read more). Troubleshooting focused on verifying whether the mails were logged/sent via EmailLog and checking /logs + /sidekiq for enqueue/sending behavior (read more), with follow-up evidence showing only the test message present in EmailLog (read more).
markvanlan drove the launch discussion for nested replies, covering enabling options, bulk actions, and reversibility (read more, read more, read more). He also clarified why threading is an admin-level decision rather than user preference (read more).
lindsey shipped and discussed the new solved “Me too” workflow, inviting ideas for admin-facing reporting and surfacing mechanisms (read more, read more).
martin explained the new anonymous/logged-in pseudogroups and reassured admins that category visibility won’t unexpectedly change when the upcoming change is enabled (read more, read more). He also landed a clarity improvement related to reporting permissions options in an ongoing announcements thread (read more).
awesomerobot handled multiple fixes and clarifications: distinguishing “Me too” from “boost” in solved contexts (read more), confirming the group owner visibility regression and opening a PR (read more), and pushing an update to the Air Theme to address reported issues (read more). He also opened a PR for a minor LLM default selection UI improvement (read more).
zogstrip fixed the grid lightbox ordering bug and linked the core PR (read more). Separately, he noted a mention-check fix revealed a new edge case affecting AI agent mentions and linked the PR tracking that nuance (read more).
nat helped diagnose translation behavior (including post-length limits) in Support (read more) and acknowledged the Norwegian locale detection mismatch, proposing prompt-level mitigations while a core prompt update is considered (read more).
chapoi provided the solved answer for disabling user cards for anonymous visitors via a site setting (read more).
dax jumped into the “Me too” launch thread to explain it wasn’t enabled on Meta, then enabled it for testing on try.discourse.org (with a brief “power went out” interruption mid-setup) (read more, read more).
Falco supported self-hosted Discourse AI sentiment users by noting an upstream model PR wasn’t merged yet and recommending pointing to his branch directly, with updated docker guidance (read more).
isaac closed the loop on the discourse-patch-triage bot workflow, detailing how to interact with it, what actions it supports, and why forks can’t receive bot commits—then followed up to raise fork write-access internally (read more, read more).
If yesterday was any indication, today’s going to be deeply threaded—just try to keep your debugging sessions shorter than the neighborhood power outages.
putty introduced a small UI tweak to hide the “votes left” line when admins simulate “unlimited” voting by setting huge limits—followed by a discussion on whether it should reappear when votes get low (theme component, threshold idea, intent/constraints — read more)
Community scaling pain points got a thoughtful deep-dive: jpishgar highlighted taxonomy breakpoints and “buried knowledge” as activity explodes (taxonomy/structure breakpoint), ice.d shared the lived reality of moderation load and owner absence (staff burden story), and ted pushed on how to design transparent moderation at scale (systems question — read more)
#Announcements
The newly announced “Me too” capability for solved topics continued to evolve via feedback: NateDhaliwal helped folks find it under Admin → Upcoming Changes (where to enable it), EricGT proposed a complementary “Works for me now” signal to reflect real-time recoveries (suggestion), and RGJ questioned whether the counter should start at 0 vs 1 and asked for visibility into who clicked it (counter + attribution — read more)
Third-party client implications of nested replies kept rolling in: nicolsdennis asked whether the Top/New/Old sort could be exposed via the existing flat topic JSON endpoint so clients can benefit without adopting nested rendering (API/sort question — read more)
Admin UX snag: EricGT reported that when opening Upcoming Changes from inbox notification URLs, the Preview link disappears, making it easy to miss the intended workflow (report — read more)
A UI regression report landed on the long-running Foundation modernization thread: NateDhaliwal noted severe overlap in the pinned state menu options (screenshot + report — read more)
A subtle signup edge case around custom user fields got a concrete fix direction: ted confirmed inconsistent sanitization between signup vs profile update, and posted a PR to align behavior (analysis — read more)
AI localization configuration: Norwegian was being detected as no while the localization setting expects nb_NO; nat confirmed the correct approach and planned prompt improvements so it respects supported languages settings (team confirmation — read more)
#Feature
Moderation ergonomics: ice.d suggested a multi-select “flag as spam/off-topic” workflow to reduce repetitive clicking and potentially enable batch handling in the review queue (moderationreview-queue) (proposal — read more)
Branding/UX for first impressions: Ethsim2 proposed a separate mobile_logo_dark (or login-specific) setting for the login/signup splash screen so admins don’t have to compromise between the unauthenticated splash and the compact logged-in header (loginux) (feature request — read more)
#Self-hosting
Security heads-up: AquaL1te shared a write-up on chaining an nginx rewrite-engine RCE with a kernel LPE, and RGJ clarified it doesn’t apply to a standard Discourse install (still useful reading for self-hosters with bespoke nginx setups) (advisory link, scope clarification — read more)
Composer mode issue confirmed as plugin-related: Hasimir reported the default_composition_mode bug goes away when disabling Custom Wizard, but noted the real constraint is “we need the Wizard” (confirmation — read more)
#Plugin
Discourse Workflow plugin got a usability upgrade: merefield added a new visual editor to make step editing less opaque, and bundled fixes for some deprecations (visual editor PR announcement — read more)
Discourse Subscriptions (official) saw a serious real-world edge case report: a Stripe PaymentIntent is allegedly created without a valid payment method in certain “expired card then update” flows, causing silent failures for end users (bug report — read more)
(More threads that also saw activity, if you’re doing a broader scan: Graceful theme discussion + pattern replacement work (Graceful); OpenID Connect username edge case (OIDC); “anonymous email” category issue persisting in 2026.5.0 (bug thread); signup URL parameter behavior (support); Custom Wizard maintenance concerns (plugin thread); plus install/how-to docs referenced by new theme components (theme install guide, themes beginner guide).)
HAWK weighed in on scalable moderation governance—emphasizing trusted people, regular comms, documented processes, and clear appeals—continuing the “breaking points as communities grow” thread in #Enterprise (reply — read more)
awesomerobot shared an important update for topic-voting admins: a new site setting Topic voting enable vote limits can remove limits entirely, reducing the need for huge “workaround” limits (while still noting the theme component’s value for mixed trust-level configurations) (vote-limit setting note — read more)
awesomerobot also addressed a design concern in the official Graceful theme: after noting the pattern’s association risk, they proposed swapping it for a different traditional Japanese pattern and linked the PR (context + PR — read more)
nat confirmed how to fix Norwegian locale mis-detection for AI localization and noted an upcoming improvement to the generic system prompt so it better respects configured supported languages (confirmation + plan — read more)
That’s a wrap: yesterday’s meta was all about making tiny UI labels do big clarity work—while “Me too” counters, vote limits, and nested reply sorting all tried to politely line up and behave.
Gamification admin UX: creating a new leaderboard wiping an older one was raised in Discourse Gamification.
Upcoming Changes metadata: optional “Feedback” and “Preview” links were clarified in Upcoming Changes.
Interesting Topics
#Announcements
martin announced that high context topic cards are now enabled by default in the Horizon theme (with a recap of signals like Solved / Voting / Assign / tags / excerpts) — read more. (Related: the original Horizon cards intro is here: read more, plus plugin refs for Discourse Solved, Topic Voting, and Assign.)
In the nested replies rollout, markvanlan explained why the team didn’t extend Post Voting to achieve the same effect, emphasizing maintainability and UX in this reply.
RGJ flagged that the Delete category action appears far more prominent than Save in the simplified category screen, creating a risky “oops” moment in this UX report.
ted questioned whether a “superfluous” toggle/button belongs in the edit bookmark dialog; chapoi ultimately removed the toggle (and linked a PR) in the follow-up (bookmarksfixed).
ted proposed showing uneditable user fields as disabled inputs (instead of hiding them) to reduce confusion; chapoi raised questions about hidden-field use cases in the discussion (user-custom-fields).
A composer edge case: placing the cursor before a horizontal rule and pressing ↓ can cause the HR to steal focus and close the emoji picker; repro + recording are in this bug report (composeremoji).
Lilly shipped improvements to Discourse Reply Button Labels, including a new setting to show the topic reply button label on mobile viewports (with notes on why post-reply labels would break the mobile footer) in this update.
#Feature
Moin suggested suppressing the “New features available!” notice on brand-new forums, arguing that everything is “new” for first-time admins in this feature request.
Ethsim2 opened a draft PR to support a separate dark-mode login/signup splash image (new splash_screen_image_dark setting) in this implementation update (loginux).
In feedback about Discourse Embedding, Falco confirmed the login friction loop (new tab → logged in → iframe still logged out until refresh) is now fixed and pointed to an update path in this reply (embedding).
#Enterprise
A cautionary scaling story: a community’s information architecture collapsed under thousands of categories, and HAWK reinforced why scalable IA (and tags) matter in this exchange (enterprise-ready).
A community member suggested nominating Discourse for the Digital Public Goods Alliance registry as both a values and marketing win, with links to the registry process in this topic.
markvanlan added strategic context around nesting and long-term maintainability, including how it relates to post-voting-style experiences, in Introducing nested replies.
That’s a wrap—may your emoji picker keep focus, your category editor keep “Delete” far away from “Save,” and your bookmarks stay blissfully toggle-free.
Nginx CVE-2026-42945—are Discourse containers vulnerable? A security question kicked off with k0tsk, darkpixlz pushed back on scope, and sam confirmed Discourse will update the base image and noted default rewrite rules aren’t vulnerable. read more, read more, read more
Discourse AI translation “not working” turned out to be a preference toggle: EricGT reported translation confusion, and Lilly/Moin helped trace it to the “show original language” preference (and the topic-level toggle tied to it). read more, read more, read more (Background referenced: Viewing localized contentread more)
Trouble deleting topics, plus nested replies edge cases: In a “why can’t I delete topics?” thread, Falco clarified the user-facing rule (only if no replies and <24h old), while follow-ups suggested Nested Replies Testing can interfere with moderation/admin actions in nested view. read more, read more, read more (Related: wrench actions failing in nested mode read more, deleted nested topic causing issues read more, permissions reference mentioned read more)
/about page group visibility: hidden groups vs overlapping visible groups: Ethsim2 documented a scenario where hiding a group section on /about didn’t prevent a user from showing up via another visible group, and asked whether that’s intended behavior. read more
Tags on the hosted Free plan: A quick clarification thread noted that tags aren’t available on the hosted free tier right now (upgrade/self-host needed). read more, read more (Related: Free plan beta announcement read more)
Theme Creator export shows a white screen (mobile-only repro so far): kairo162 reported a broken export flow, and Lilly confirmed it reproduces on mobile (desktop OK), leaving the page stuck blank. read more, read more
How do you close full-screen reactions on mobile?Lilly described getting trapped in a full-screen reactions UI on iPhone/Chrome with no obvious dismissal except reloading the site—hard to reproduce, but disruptive. read more
Flags requiring a message can crash and fail to submit (TypeError): cuo_wu reported an uncaught error when submitting require_message flags (e.g. “Something else”), with a console stack trace and root-cause notes. read more
Nested mode breaks post admin “wrench” actions (and related deletion/recovery pain): Lilly reported that wrench actions appear but fail silently in nested view, requiring switching back to flat mode for moderation actions. read more (Related nested/deletion symptoms: read more, read more)
Older WordPress-embedded topics missing noindex + canonical: Thiago_Mobilon reported older embed-created topics losing the X-Robots-Tag: noindex header and canonical back to WordPress; angus asked targeted questions to narrow down what changed. read more, read more (Possibly related embed/nested behavior observed elsewhere: read more)
Upcoming plugin name clash: “Workflow”: In a naming dispute, j.jaffeux advised third-party developers to avoid “Discourse X” style names to reduce collisions (suggesting explicit vendor namespaces), with the plugin author disagreeing. read more, read more
Dev bootstrap: creating the first admin without email + what the dev toolbar does: Ethsim2 summarized a clearer mental model for admin creation in dev (d/boot_dev --init, MailHog, rails admin:create), and Moin pointed out the vertical developer toolbar and its toggles (safe mode, verbose localization, outlets, etc.). read more, read more (Toolbar intro: read more)
#Theme component
Quote Callouts: updates to history rendering, excerpts, and shortcut: Arkshine shipped improvements so callouts render in revision history, onebox excerpts strip the marker, and the keyboard shortcut changed to ctrl/⌘ + alt + c. read more
#Feature
Change the “More…” reactions icon when any-emoji reactions are enabled: A feature request proposes replacing the gray smiley used as the “More…” affordance in reactions, continuing the earlier thread about reactions UX. read more (Context: read more)
martin shipped an #Announcements guide to streamline creating Events categories—automatically enabling key calendar/event settings and wiring the category into the events calendar configuration. read more
sam responded to the Nginx CVE thread, saying Discourse will update the base image and noting default rewrite rules aren’t vulnerable. read more
Falco clarified topic-deletion limits for regular users (no replies, under 24h), and separately flagged what looks like a bug path involving deleted nested topics (planning to investigate). read more, read more
j.jaffeux weighed in on a plugin naming conflict, recommending developers avoid “Discourse X” naming to reduce clashes and adopt explicit namespaces. read more
Yesterday’s vibe check: we made event planning easier while simultaneously discovering that nested replies can hide the moderation wrench—so if your tools go missing, try “flat mode” before you start blaming Nginx. (See: events category setup, nested wrench issue, Nginx CVE thread)
Eviepayne asked for a declarative way to persist set_real_ip_from / real_ip_header X-Forwarded-For inside Discourse’s container Nginx when running behind HAProxy, after seeing client IPs collapse to the load balancer and triggering 429s (unsupported-installnginx). The discussion distinguishes host-vs-container nginx concerns and lands on using replace in app.yml to survive rebuilds. read more
Key context: why the load balancer IP caused trouble, suggested replace injection snippet
In a permissions/regression-style support question, a user reported they can’t delete topics anymore and is awaiting a fix; Prempavee closed with appreciation while waiting. read more
TroLLoBlogger documented a crash after attempting to update via the admin UI, then recovered by rebuilding from the terminal (./launcher rebuild app). The error log points to a plugin asset compilation timeout, and the thread pivots into practical guidance on update workflow and resource constraints. read more
Notable replies: Lilly on avoiding admin-UI updates for self-hosting read more, and Falco requesting server specs to understand the 120s build timeout read more
In an older-but-still-active bug report, Moin pointed to a fix PR for the “Thanks for Contributing” popup still appearing after the intended limit had been passed. read more
sniper756 reported that, on mobile browsers, the admin tag-group management page renders incorrectly (noted alongside “Modernize Foundation theme”), with a screenshot showing layout issues. read more
#Community Building
A discussion on using Discourse for small town government sites focused less on software capability and more on institutional realities: moderation ownership, what counts as “official,” authentication, security, and user expectations. Ethsim2 emphasized that the human/process side is the hard part read more, while NateDhaliwal highlighted privacy implications—whoever hosts has “absolute access.” read more
Ethsim2 proposed clearer behavior for /about group listings when “hidden groups” and “extra groups” overlap, including the idea of a boolean site setting like about_page_exclude_hidden_group_members to prevent unintended exposure through alternate group sections (groupsabout-page). read more
The back-and-forth covers what each setting actually controls and where group visibility rules apply: Nate’s interpretation of the two settings, Ethsim’s reproduction/edge case, and the nuance that group interaction settings may still affect /about rendering read more
mcwumbly jumped into a heated hosted-pricing/support thread to clarify that Starter has been deprioritized on the pricing page but should still be available from Free upgrades, then began troubleshooting where the upgrade flow was being triggered. See: initial clarification and question read more, confirmation it “should” appear and plan to investigate read more, an attempted quick fix read more, and the final “hang tight” update pending deeper review read more.
Falco investigated a self-host update crash tied to plugin asset compilation timeouts, narrowing suspicion to the plugin build exceeding the two-minute limit and requesting server specs to confirm whether resources were the bottleneck. read more
May all your upgrade buttons reveal the plan you actually want—and may every ./launcher rebuild app finish before the timeout gremlins notice.
greymont — 3 posts, 4 likes (closing the loop on a hosted signup/verification confusion in read more)
david — 1 post, 4 likes (clarifying how prebuilt assets are distributed and where plugin builds can bite in read more)
nat — 2 posts, 4 likes (sharing quick updates on bulk-tagging + AI locale detection in read more and read more)
Richie — 2 posts, 3 likes (confirming a theme-component update after a PR merge in read more)
A lot of today’s “hands-on admin reality” clustered around updates + nested replies behavior (see read more, read more, and read more).
Interesting Topics
#Announcements
Discourse Events got a usability boost: recurring-event RSVPs now let members choose “this occurrence only” vs “this and all future occurrences”, helping organizers interpret attendance and reminders more accurately (calendar-and-event) — read more
The Horizon theme now enables “high context” topic cards by default, and feedback immediately surfaced around making that configurability more granular (ideally per-category) (horizon-theme) — read more
A request for group notifications for upcoming events resurfaced in the context of the new RSVP workflow (and the historic workaround of group PMs) — read more
“Upcoming Changes” admin UI got a small-but-useful confirmation: the Preview button now appears with the expected filtered URL, helping admins validate which flags they’re previewing (upcoming-changes) — read more
Self-hosters reported update crashes and rebuild pain, with discussion centering on resource constraints (notably single vCPU setups) and how non-core plugins can force asset builds that tip low-resource servers over — read more
Nested replies continues to shake out: the post-admin wrench actions can fail in nested mode, and the emerging guidance is to switch to flat mode as a workaround while fixes land (nested-replies) — read more
Related nested-replies fallout: recovery actions for deleted posts were discussed alongside references to fixes in progress (nested-replies) — read more
A report flagged a TypeError when submitting certain custom flag flows (“require_message” flags), and the team invited reproduction details (and even PRs) — read more
AI plugin link-click behavior was reported as “no response” in a Chinese-language bug report (ai) — read more
A brand-new hosted admin hit a confusing “Oops! page doesn’t exist or is private” during email verification; the root cause appeared to be testing while still authenticated as admin in another window/session — read more
“Change Ownership” appeared to do nothing for an admin using nested replies; switching to flat mode resolved it, reinforcing that nested mode is still blocking some post-admin tooling (nested-replies) — read more
A moderation best-practices thread asked whether Discourse ships with a default profanity/abuse list; the answer highlighted Watched Words as the built-in mechanism, but without a default list (you bring/import your own) — read more
A UX-writing/editor question: users want an easy “post-number mention” shortcut (e.g. #123) to link to a reply without quoting; today’s best workaround remains copying the post link, or implementing custom behavior via theme component — read more
#Feature
A feature request proposed adding Group selection when bulk converting Topics → Messages, enabling a smoother “convert a category into a group inbox” workflow rather than ending up with messages limited to current participants — read more
A smaller admin-cleanup request asked for auto-purging silenced spam accounts with no posts after X days; discussion pointed back to existing cleanup tools and the “delete and block” spam flow — read more
#Theme component
“Normalized Profile Links” got maintenance love: Lilly’s PR fixed array deprecations and was merged, with confirmations that updates worked cleanly on live sites — read more
“Quote Callouts” received a CSS fix for <code> styling in callout titles (alignment/visual consistency, especially noticeable with certain fonts and CJK text sizing) — read more
#Self-hosting
A self-host admin lockout scenario: disabling Discourse ID before establishing local admin access can leave “no login methods configured”; recovery guidance centered around creating an admin via Rails (rake admin:create) — read more
lindsey shipped an Events quality-of-life improvement: recurring RSVPs now distinguish between “next occurrence only” and “all future occurrences,” reducing reminder noise and clarifying attendance for organizers (calendar-and-event) — read more. She also pointed older support threads to the new behavior to close the loop (see follow-up linking in read more).
Falco spent time in the trenches on nested replies + admin tooling: clarifying expected behavior boundaries and pointing bug reports into the right funnel in read more, then linking an in-flight fix for wrench-action failures in read more. He also referenced a fix for deleted-post recovery issues in read more, and helped diagnose low-resource update failures in read more.
david weighed in on update crash analysis, clarifying how prebuilt assets should be pulled even during web-based updates, and narrowing the likely culprit to non-core plugin asset build time rather than core itself — read more.
awesomerobot pushed forward on UX consistency: agreeing that the “More…” emoji trigger should be consistent when allowing any emoji for reactions (reactions) and working on a fix — read more. He also recommended the “delete and block” spam flow as best practice while discussing account noise cleanup — read more — and incorporated feedback into solved-status counting behavior — read more.
nat provided two quick “this is landing” style updates: noting a returning capability around bulk-tagging management — read more — and confirming that the default AI locale detection agent now includes better language-code handling for Norwegian/Greek (ai) — read more.
HAWK expanded on moderation tooling: confirming there is no default blocked-word list, but detailing the breadth of Watched Words actions (block/require approval/flag/censor/etc.) and import/export capabilities — read more. He also reassured Enterprise users that calendar/events work is active and video conferencing is on the roadmap — read more.
featheredtoast responded to concerns about the discourse/discourse Docker image update cadence, explaining a slowdown was an unintended ESR-branch side effect and would be corrected — read more.
sam acknowledged a bug report around a flag-submission TypeError and invited more info (and potential community PRs) — read more.
Yesterday’s vibes: if your recurring RSVPs are finally behaving and your 1vCPU server survived an update cycle, you’ve basically cleared Discourse hard mode.
In the last 24 hours (2026-05-18 → 2026-05-19), Meta saw 87 new posts spread across 16 new topics, including multiple product announcements like starring AI conversations (read more) and a docked composer for AI bot chats (read more), plus an intermediate security release (read more).
Shared progress on adopting Discourse for small-town government (read more)
Interesting Topics
#Announcements
sam announced a quality-of-life improvement for Discourse AI: you can now star common AI bot conversations so they remain pinned at the top of your chat list (toggle coming under upcoming changes) — a small tweak aimed at making AI chats feel more “messaging-like.” (read more)
keegan unveiled a new docked composer for aiai-bot conversations, keeping the input persistently available at the bottom of the topic so you don’t have to click Reply for every turn in a back-and-forth chat. (read more)
david posted May 19th 2026 intermediate releases, pointing admins to critical/security update streams and changelogs for multiple supported lines. (read more)
Follow-up on Solved improvements (“Me too” for reported issues): admins asked whether the “Me too (%{count})” text is customizable, and awesomerobot confirmed it’s safe to change via site text. (read more, read more)
The Events team published an Office Hours replay featuring a deep dive on MCP + Data Explorer, demonstrating how AI assistants can generate queries and translate community analytics into natural language workflows. (read more)
chapoi introduced an upcoming-change flag for a Form checkbox redesign (FormKit), visually hiding the native checkbox while preserving functionality and enabling smoother styling/animation. (read more)
“Where did the tag info button go?” got an official answer: it was removed after adding a dedicated tag settings page, and awesomerobot opened a PR to bring it back with a refreshed banner aligned to category banner styles. (read more)
#Theme / #Theme_component
The official Horizon Theme thread saw renewed attention on small-screen ergonomics: users reported odd “suggested threads” rendering on mobile emulation and raised concerns that the pinned category/tag navigation consumes too much vertical space; Lilly traced the behavior to navigation-container and suggested CSS workarounds for button/text layout. (read more, read more, read more)
Discourse Mermaid users asked about the cadence for updating the underlying mermaid.js library (specifically to reach newer chart types like Wardley Mapping), and Falco encouraged contributors to submit a PR. (read more, read more)
#Migration
A large phpBB → Discourse migration prompted performance troubleshooting: the importer was running ~440–450 items/min while CPU stayed low, with Redis timeouts and skipped posts appearing in logs; discussion focused on I/O/DB bottlenecks, Redis diagnostics, and the tradeoffs of bulk importers. (read more, read more, read more)
A self-hosting admin reported that backups fail with PostgreSQL 18 due to a pg_dump client/server version mismatch (container has pg_dump 15.x vs server 18.3), raising the practical question of whether to align Postgres versions, upgrade the client, or segregate DB clusters. (read more, read more, read more)
A report of Internal Server Error when approving new users (with OAuth2 auth) focused on where to look first: /logs, browser console, and approval workflow diagnostics. (read more, read more)
A tricky bug hunt continued for a TypeError when submitting “require message” flags, where postActionFor can return undefined in certain “nested replies” situations; repro steps narrowed in on flagging in nested mode and then selecting another notify-type reason. (read more, read more, read more)
A mobile-only issue reported a 404 at the bottom of a nested replies testing topic, suspected to relate to recent nested read-state behavior; additional context connected to broader “topic list state / old topics showing” weirdness. (read more, read more, read more)
Voting UI regression watch: users can’t view who voted on a closed topic when topic voting is enabled, with both the UI and API returning empty voter lists—raising concern that an older bug may have resurfaced. (read more)
sam shipped Star common AI conversations to keep frequently used bot chats surfaced (#Announcements, ai) (read more).
keegan announced the docked composer for AI bot conversations, now available behind an upcoming-changes flag (#Announcements, ai-bot) (read more).
david published May 19th 2026 intermediate releases with pointers to the official changelog streams (#Announcements, release-notes) (read more).
Danielle posted the Office Hours replay: a guided walkthrough of MCP + Data Explorer and how AI can assist with community analytics workflows (events) (read more).
chapoi introduced an upcoming-change flag for the Form checkbox redesign (ux) (read more).
awesomerobot confirmed site text customization for “Me too” in Solved’s reported-issue UX (#Announcements, solved) (read more), noted that brand new sites won’t be spammed by historical “New features available!” notices (#Feature) (read more), and opened work to restore the tag info button UX (ux) (read more).
Falco connected a mobile nested-replies 404 report to recent nested read-state changes (bug, nested-replies) (read more), encouraged contributions to Discourse Mermaid update cadence (#Theme_component) (read more), and shared improved custom SVG splash examples (#Announcements, splash) (read more).
mae shared that Discourse was submitted to the Digital Public Goods Alliance registry (a step toward broader public-interest adoption) (General) (read more).
nat merged a fix after adding specs for a language-code-related bug, thanking the contributor for the PR (bug, translation) (read more).
isaac marked updates to @patch-triage, enabling review/rebase/commit requests and improving fork behavior (development) (read more).
Yesterday’s theme was unmistakable: Meta got a bit more “chatty,” with AI conversations now star-able and the bot composer docking itself like it owns the place.
In the last 24 hours, meta saw 63 new posts across 16 new topics—with conversations spanning everything from open web values to chat quirks and migration performance. Activity clustered around threads like the blog post on staying browser-first (read more), the #Migration importer tuning discussion (read more), and a handful of fresh bug / ux reports (e.g. read more, read more).
ducks argued that Discourse won’t force mobile users into an app, positioning Discourse as explicitly pro–open web in contrast to app-first lock-in trends (read more). NateDhaliwal also flagged a housekeeping detail about the topic author attribution (read more).
A deep-dive on speeding up phpBB → Discourse imports focused on database I/O bottlenecks, with cocococosti recommending pragmatic tactics like adding indexes/views and tuning DB settings—and teasing “new tooling” aimed at dramatically faster migrations (read more). The thread also includes real-world importer failure logs from an ongoing run, useful for diagnosing repeatable edge cases (read more).
Richie requested small consistency tweaks for the expanded search UI—specifically the casing of dropdown options and removing an unnecessary ellipsis—while NateDhaliwal noted that lower-case options read better in-sentence (“Where topics are closed…”) and agreed the other dropdown likely needs standardizing (read more, read more, read more).
RGJ spotted an auditability gap: restoring (undeleting) a post does not create a staff action log entry, unlike deleting/restoring topics—making it harder to see who restored content without digging into system logs (read more).
#Announcements
Building on the Events + Calendar work, martin explored whether Discourse could auto-enable an upcoming-events sidebar block when the “right sidebar blocks” component is present, and nathank proposed a more unified Events category view combining calendar-style “upcoming” visibility with a topic list (read more, read more).
In the “Me Too” + Solved improvements thread, EricGT suggested making “Me Too in closed topics” a customization option, and floated an approach where “Me Too” on a closed issue could prompt creation of a new topic after AI checks for duplicates (read more). The discussion also branched into whether the button should appear in deleted/closed contexts, referencing a related layout thread (read more).
davidb announced an opt-in layout tweak for sites using Discourse Reactions: moving reply count to the right to keep reactions alignment consistent, enabled via upcoming changes as Enable new post reply count position (read more).
For WooCommerce + WP Discourse setups, angus clarified that usernames don’t need to remain in sync: DiscourseConnect maps accounts via external_id (WP user id), not the WordPress username—so letting users change their Discourse username shouldn’t break SSO matching (read more).
A WordPress embed SEO issue: older embedded topics were missing x-robots noindex and canonical tags, and angus narrowed it to missing topic_embed records required for canonical-url-for-embeds—while Thiago_Mobilon argued for a more robust category-level “noindex all topics” checkbox to avoid reliance on embed records (read more, read more).
A site admin asked how to export all site data (not just a personal archive). NateDhaliwal pointed them toward Admin → Backups for a full instance backup suitable for upgrades and server moves (read more, read more).
A theme-component update attempt hit about.json does not exist during import; responders suspected ZIP structure issues (parent directory nesting differences across OS zip tools). Moin demonstrated that re-zipping with the right folder structure fixed installation (read more, read more).
Postgres tooling friction surfaced again: a self-host reported backups failing with Postgres 18, prompting advice to install matching postgres client packages inside the container and then persist that setup via app.yml rebuild hooks—plus a pointer to prior related discussions about upgrading the postgres client (read more, read more, read more).
A small but impactful paper-cut: the “update available” email included a broken release-notes link due to a missing v prefix in the changelog URL (read more).
Pricing anxiety continued: a user tracked visible shifts on the Discourse pricing page and reported previously-available features now appearing “Upgrade to unlock” on a recently created community, connecting their concerns to earlier free-plan beta discussion and a “no starter option” thread (read more, read more, read more).
Moin reported a Unicode edge case in emoji autocomplete (notably with ü) where selection via “more” fails to replace typed text; in chat it escalates into a console error (emojiTermStart null) and broken selection behavior (read more). The report also cross-referenced an older “revision despite grace period” observation triggered while reproducing (read more).
Starred chat UX bug: Shauny described how the Chat “Back” button forgets you were in Starred, always returning to Channels/DMs, undermining the time-saving intent of starring channels—and pointed to the original “star your favorite chat channels” feature announcement for context (read more, read more).
A tag-icons regression hunt: AquaL1te noticed tag icons rendering inconsistently (sidebar OK, topics not), and Lilly hypothesized an underscore-vs-dash slug mismatch introduced by a component change—suggesting updating slugs in config rather than renaming tags (read more, read more).
ducks published the day’s headline blog thread asserting Discourse’s browser-first stance: “Discourse Will Never Block You From the Web” (read more).
cocococosti advised on accelerating phpBB imports by targeting the typical bottleneck—database I/O—and recommended indexing/views/tuning to make repeated imports less painful, while hinting at upcoming faster migration tooling (read more).
martin chimed in on Events category creation UX, noting a potentially valuable integration point with right-sidebar blocks and proposing investigation into enabling an events block when the relevant component is installed (read more).
davidb announced an upcoming change for sites using Reactions: an opt-in reply-count position adjustment to improve alignment consistency, enabled via the upcoming-changes UI (read more).
markvanlan followed up on a reported 404 at the bottom of a nested replies testing page, noting they couldn’t reproduce at the time but would keep an eye out for recurrence (read more).
Yesterday’s energy was all about staying on the open web—even as folks simultaneously shipped new UI polish and dreamed up native app experiences—so however you’re browsing today, at least you’ve got options.
Biggest “headline” drop: a new reading-first “simple mode” for Doc Categories landed as an announcement — read more
Interesting Topics
#Announcements
A new “simple mode” for Doc Categories (reading-first docs, replies behind a toggle).saquetim introduced a documentation-shaped topic experience in #Announcements with doc-categories, including auto-expanding when linking directly to replies and a simplified “Updated” list column — read more (background: Discourse Doc Categories plugin, and a real-world “oops, I missed earlier replies” example: Backup from command line)
iOS PWA nested replies layout gap (username/profile spacing).kairo162 reported an odd spacing issue in nested replies on iOS PWA, and markvanlan said they’ll dig in — read more (team acknowledgment: read more; related testing thread: Nested Replies testing)
Horizon theme header breaks on mid-size/folding-phone viewports (640–767px).nathank shared screenshots where the search/avatar vanish at a specific width range, with Lilly offering a CSS workaround and tying it to viewport-based mode switching — read more (CSS investigation: read more; related discussion on mobile/desktop switching: read more)
Core reactions “+1 hover popover” stuck loading due to URL encoding.PortChange traced the issue to reaction_value=+1 being interpreted as a space, and zogstrip pointed to a fix; the reporter confirmed it was working quickly after — read more (fix noted: read more)
Mobile UX: browser back button while reaction list modal is open.Canapin reported the reaction list staying open and the page jumping to the footer on Chrome Android; Moin connected it to earlier reaction-modal work — read more (related earlier posts: read more, read more)
Emoji autocomplete + special characters (e.g., “ü”).renato linked a proposed fix PR for emoji autocomplete behavior involving special characters — read more
“Users for approval” notification claims approvals exist, but the queue is empty.Christian_Bailey confirmed they’re hitting the same “phantom approvals” bug (and that they’re the only admin, ruling out someone else approving) — read more
Flagged post restoration is confusing when a “reason given” blocks restore.zogstrip credited a community find and pointed to a fix-incoming PR — read more
Hosted upgrade UX: “Starter” plan not selectable (only Pro shown), partially fixed.Crown1 reported that Starter details appeared but couldn’t be selected from some upgrade surfaces; later confirmed “Manage Account” now works, and rishabh explained the underlying recommendation logic issue (Free/Starter staff-limit parity) and promised a fix — read more (confirmation it worked via Manage Account: read more; team follow-up: read more)
Free plan / pricing structure confusion: category types and missing feature-matrix rows.HAWK clarified the Free plan is still new/being tuned, noted transactional-email limits (and an “email in” oversight that got reverted), and agreed feedback about the feature matrix and category-type changes is fair — read more (category type concerns: read more; follow-up: read more)
Backups failing with Postgres 18: solved with package changes. A backup failure thread ended with a working solution flow (installing/upgrading Postgres packages), plus pfaffman recommending pinning client packages and moving commands into app.yml — read more (follow-up advice: read more)
View count doesn’t increment on one topic (others are fine).Prempavee asked why a single topic isn’t counting views; NateDhaliwal asked for specifics and suggested testing safe mode to rule out theme/component interference — read more (triage questions: read more)
A native iOS+Android Discourse client app (preview) and what it needs to work broadly. In show-and-tell, nicolsdennis explained the app isn’t store-published yet, and that some features depend on an accompanying plugin + configuration like Allowed user API auth redirects — read more (implementation details: read more)
#Plugin
Chat UX: floating chat + browser history causing confusion.sam confirmed a reported “back button” issue is specific to floating chat mode and argued floating chat shouldn’t participate in page history; zogstrip noted it should improve once a fix is merged — read more (status update: read more)
Discourse Automation: can an auto-reply mention the user? A question in the official automation thread asks whether automated replies can @mention the newly-posting user (useful for “first post” onboarding patterns) — read more
#Theme component
Featured Topics component: possible collision with a Pavilion dropdown header.TestingRico reported that previewing the Featured Topics component seems to make the Pavilion dropdown header disappear, asking if it’s known and how to keep both working — read more (related component referenced: pavilion dropdown header)
WordPress embeds: old embedded topics missing x-robots: noindex and canonical tags.angus explained canonical URLs should work, but missing embed records imply some external deletion/operation; Thiago_Mobilon argued category-level indexing control should be native and that republishing isn’t viable at large scale — read more (SEO feature request argument: read more)
Pointed to an upcoming fix for chat floating-mode back button quirks — read more
Posted a hopeful fix direction for Googlebot 404s from page number URLs — read more
Confirmed a UX tweak (“Auto silence first post regex” description should link to watched words) is pending merge — read more
Investigated the hard-to-repro full screen reactions close behavior on mobile — read more
Confirmed the encoding-related “+1” reaction hover popover bug would be fixed — read more
Flagged-post restore workflow: acknowledged a find and linked the fix path — read more
david responded on the folding-device/mobile-mode discussion, noting Discourse design work assumes viewport-based mobile mode remains enabled and that Horizon should be fixed without toggling that setting — read more
HAWK engaged on hosted-plan expectations and market-fit iteration (Free plan, email constraints, category type changes), and took feedback to raise internally — read more (follow-up: read more); also continued the growth/process conversation in #Enterprise — read more
markvanlan acknowledged and agreed to investigate the iOS PWA nested replies spacing bug report — read more
featheredtoast updated self-hosters using the official Docker image that the next build should restore expected upgrade behavior (and later confirmed it was back) — read more
sam clarified the floating chat context for a back-button report and suggested floating-mode chat should avoid participating in browser history — read more
rishabh apologized for the hosted-plan upgrade confusion and explained why Starter was incorrectly skipped by the recommender, promising an update once fixed — read more
renato linked a fix PR for emoji autocomplete issues involving special characters — read more
supermathie asked for more details on a Chinese-language support post about the review queue (“what reason was given for the flag?”) — read more
Yesterday’s vibe: between simple-mode docs, folding-phone headers, and reaction popovers that don’t like plus signs, Meta was basically a usability escape room—now with better lighting and fewer hidden toggles.
Discourse published “Discourse is Agent Ready”, sparking a practical debate about whether communities should welcome or block AI crawlers; awesomerobot framed it as “depends on whether your forum is support vs social,” while others raised GDPR/LGPD and scraping concerns in the companion thread read more.
A potentially serious moderation workflow issue emerged: admins report the “Approve” button on a user profile can fail, leaving users unapproved yet missing from the review queue; Moin’s repro and Ed_S’s “me too” add weight to the report in Approval button on user’s profile doesn’t work (closely related to the “phantom approvals” thread here).
A copy/paste edge case: syntax highlighting colours don’t reliably carry over when copying code blocks into external clients; supermathie showed that the codeblock “Copy” button yields plain text, while selecting rendered HTML can preserve formatting—then the thread narrowed to browser/client behavior differences in this bug report.
iOS PWA composer UX: when text size is set smaller, the Reply button can disappear in the iOS PWA (while Safari works), with a fresh report and screen recordings in Reply button not visible when composing on iOS PWA (noting a possibly related earlier thread here).
Foundation theming regression watch: after a rebuild/update, some sites saw very small font sizes on category-homepage layouts, and Lilly traced it to “Modernize Foundation theme” and offered an SCSS variable fix in 刚才rebuild更新后，首页字体变的很小 (foundation-theme).
Threaded/nested UI parity request: in nested/threaded view, admins can’t easily add notes to action-log entries, even though it’s possible in the flat view—feature request and screenshots in Can’t add note to admin action in threaded view (nested-replies).
Free-tier clarity: awesomerobot explained what “minimum activity” means for hosted free sites—roughly a couple users + a handful of topics in the first ~3 weeks, with notices at ~3 and ~6 weeks—while confirming the embed setup looked correct in What is the minimum required activity? (free-plan).
Mobile navigation oddity (Chinese support): a report suggests iOS Chrome’s back button can behave inconsistently on Meta—sometimes returning to an older page instead of the immediate prior view—in Chrome浏览器返回键的bug.
Attribution vs simplification: Lilly asked for “simple mode” to be disabled (or contributors restored) in the Wiki category, arguing the current presentation can unintentionally erase visible credit for non-staff authors in this feedback topic.
Meta’s own digest mechanics were discussed again in the long-running “daily summary” thread, including how top-user callouts and references get presented in Daily Summary Feedback.
#Announcements
The nested replies rollout continues to spur customization discussion: one thread revisited how to handle reply-label language/translation tradeoffs, pointing from the nested replies announcement toward a dedicated labels solution in Introducing nested replies (see also the linked labels topic here).
#Plugin
In topic-voting, admins reported the familiar big “Vote” button has become a small arrow that users overlook, prompting requests for UI reversal or customization guidance in Discourse Topic Voting.
awesomerobot engaged in the “agent-ready” discussion, outlining when AI visibility might be a net positive (support/answers) versus harmful (social community goals) in this reply, and also cautioned that blocking scrapers isn’t foolproof if content is public in this follow-up.
They also clarified free-tier inactivity expectations and timelines in the free-plan support thread.
supermathie broke down why syntax highlighting colours don’t always paste into other apps—distinguishing the Discourse “copy button” vs selecting rendered HTML—in this diagnostic post, adding additional troubleshooting steps in this follow-up, and explaining Firefox vs Chromium style-inlining differences in this technical note.
techAPJ confirmed how the “help answer similar questions” PM chooses suggested topics (and how “limit suggested to category” affects it) in the solved reminders thread.
If yesterday’s Meta felt like a mashup of AI crawler ethics, mysteriously tiny fonts, and haunted approval queues, that’s because it absolutely was.
In the last 24 hours, Meta saw 36 new posts across 9 new topics, with most of the churn concentrated around bug triage, permission/UX refinement, and a couple of theme-component releases (e.g. read more, read more, read more).
wenqin shared cross-cultural and generational feedback on topic reply order and discoverability for older users, suggesting admins should be able to choose a default “newest-first” view while still allowing per-user overrides (read more); jack222 echoed the friction of landing at the bottom and needing to scroll to understand context (read more).
Ethsim2 proposed splitting the broad permission delete_all_posts_and_topics_allowed_groups into own-content vs global-content capabilities to better support least-privilege role design (especially around deletion footprint vs moderation power) (read more), then dug into how PostGuardian checks differ for “seeing deleted posts” vs “seeing deleted post bodies” (read more).
Continuing the permissions thread, Ethsim2 asked for clearer wording on “delete removed posts after” to distinguish the author-facing removal action from later automatic cleanup, and to clarify what staff/regular users can see during the window (read more).
After “update all”, ice.d hit an intermittent “forum issues” situation that looked like a stale-tab vs rebuilt-asset mismatch; Falco asked to check /logs (read more), Ethsim2 identified missing Propshaft chunks and connected a separate backend exception to the bundled reactions regression (read more), and pfaffman advised a command-line rebuild (read more).
Ryan_Hyer reported broken infinite scroll in the user directory when a specific theme component is enabled, with loadMore failing because this.model is undefined (tagged user-card-directory) (read more).
cuo_wu found a 500 error when enabling both nested-replies and post-voting, triggering undefined method 'post_number' for nil in NestedReplies::PostTreeSerializer (read more); Lilly suggested the two features likely aren’t compatible yet, linking to prior nested replies discussion (read more).
Ethsim2 reported a NameError: ReactionsSerializerHelpers during topic load on 2026.5.0-latest.1, pointing at a bundled reactions serialization path (read more) and noting they opened a PR to move/load the helper earlier in the chain (read more).
A longer-running issue about copy/paste of syntax highlighting (where destinations that support coloured text still lose the colours) picked up further investigation ideas from rokejulianlockhart, including looking for relevant Chromium source references (read more).
The “Approval button on user’s profile doesn’t work” thread deepened: Moin hypothesized the issue may relate to users being reviewed before (and not being reviewable again in the expected way), sharing analysis around suspect_user and reviewable states (read more); Ed_S later found similar evidence and tied it to possible historical ID reuse / deleted-user artifacts (read more), also referencing an older related report (read more).
In a Chinese-language support thread, bird asked why the Like button wasn’t visible; supermathie found Likes are suppressed in categories using Topic Voting and linked the canonical explanation from the voting plugin docs (read more). The thread also referenced how to re-enable both Like + Vote (read more) and noted enabling reactions can make a Like-like control appear in voting contexts (read more). bird ultimately confirmed success via a CSS component workaround (read more).
A solved report about a Chrome back-button bug appears to be language/translation dependent: NateDhaliwal suspected a translation mismatch (“theme” vs “topic”) (read more), and jack222 observed the issue disappears when switching UI language to English (read more).
A moderation-permissions question asked whether category moderators can be prevented from reviewing flags on their own posts; Ethsim2 reasoned through role trade-offs (TL4 vs category moderator) and emphasized that theme components shouldn’t be treated as security boundaries (read more).
#Theme component
mcmcclur published “KaTeX titles”, a theme component that typesets math in topic titles using KaTeX (with notes about MathJax conflicts) (read more). The motivation references prior requests for math in titles and the constraints around title formatting (read more, read more).
The long-running “Discourse Indented Subcategories” theme component thread saw a post removed by its author, a small but notable moderation/maintenance blip in an actively used sidebar-related component discussion (sidebar) (read more).
#Plugin
In the official Discourse Calendar (and Event) topic, nathank noted an idea would make an excellent Contribute > UX topic, and mentioned it’s timely given a recently implemented edit button on the preview (read more).
#Community Building
A “New community” thread received a firm but constructive reminder from dax: Meta isn’t for running functional tests or promoting trial sites; use the demo instead, and avoid unsolicited PM outreach for testing workflows (read more).
supermathie diagnosed why Likes were missing in a Voting-enabled context (Likes are automatically disabled), pointed to the relevant canonical docs, and suggested possible paths forward (CSS workaround / Reactions) in the support thread “请问一下点赞按钮如何显示” (read more).
Falco responded to “Bug? forum issues” by directing the reporter to inspect /logs as the first step in narrowing down the incident after an update (read more).
dax clarified appropriate usage of Meta in “New community,” steering testing to try.discourse.org and discouraging direct PMs for test-related outreach (read more).
That’s all for today—yesterday’s Meta energy was basically “votes ate my likes, nested replies fought post voting, and KaTeX tried to sneak math into titles anyway.”
