What drove the chatter: reactions UX polish and regressions dominated the day (read more), alongside tag visibility/cleanup discussions (read more) and the ongoing push to make Discourse embedding feel like a first-class comment system (read more).
Interesting Topics
#Feature
brendantheclayboy and others pushed back on the reactions UX change, while chapoi shared an improved “all reactions” panel that supports scrolling/loading more—with remaining layout tweaks in flight (read more).
A thoughtful feature idea asked whether Discourse could support delayed/self-approval posting (or approval by trusted friends) for users dealing with episodic unwellness, to prevent “write on a bad day, regret on a good day” moments (read more).
Feedback from a real-world deployment praised the redesigned embedded comment composer (especially on mobile), while flagging a couple UI bugs and suggesting testing an “always-open” composer to increase engagement (read more).
#Announcements
Site owners experimenting with the “full comment system” embed reported confusing “half-logged-in” behavior and Safari-specific prompts; Falco clarified the logged-out mode and pointed to custom integration issues (notably with Ghost + DoG setups) (read more).
A serious performance report suggested chat replies can silently create thread tracking that bloats over time, leading to Pitchfork timeouts and chat load failures on long-history sites; the report includes query traces and suspected hotspots (read more).
Mobile moderation UX got a detailed teardown: misaligned header elements, cramped spacing, and oversized inputs on reviewables—prompting a cleanup pass and PR from core (read more).
On mobile, the post revision/history modal shows every button label truncated; the suggestion: allow two-line labels and reclaim unused top padding for a more usable layout (read more).
#Site feedback
Meta’s tag taxonomy came under the microscope again: several “restricted” or staff-ish tags appear in the public tag picker, raising questions about what “restricted” should mean in the UX and whether tag visibility should follow category permissions more strictly (read more).
Meta’s ongoing redesign/theme rebuild continues: icon work is being folded into the Lucide Icons component, and small spacing/polish oddities (like upvote container proportions) were spotted and discussed as the new UI settles (read more).
A micro-meta improvement request proposed introducing a dedicated tag for revision/history discussions; the community quickly converged on a concrete tag choice to help future searchability (now using revisions) (read more).
#Plugin
Communiteq published Communiteq Power Tools, bundling a grab-bag of admin enhancements into a single maintained plugin (including exposing “hidden” settings to reduce Rails console dependency), with screenshots and an open invitation for suggestions (read more).
The Discourse Private Replies plugin hit friction with the newer category settings UI; maintainers noted multiple plugins were “caught off guard” and shared a temporary workaround involving disabling a specific upcoming change setting (read more).
#Theme component
The Social Share theme component got a PR to allow configuring popup dimensions (or switching behavior), addressing the hard-coded default being too small for many share widgets (read more).
Composer Button Bonanza sparked a useful “docs vs behavior” moment: keyboard shortcuts weren’t firing due to a documentation mistake about the shortcut format (use shift+k, not shift-k) (read more).
A recurring question about automating Discourse access via Python resurfaced: discussion centered on public endpoints vs API keys, ToS responsibility, and the practical difference between “exploration/dev tooling” and “automation/scraping”—with core folks reiterating rate limits and intent matter (read more).
A user trying to enable an AdSense trust-level control discovered it’s a hidden site setting; guidance covered how to set it via Rails console and what “normal” console exit behavior looks like (read more).
chapoi shepherded the reactions UX fix forward, sharing a “live on Meta” preview and confirming the new placement standard (“all displays on the left”) in the reactions rework thread (read more), and also confirmed the “keep reactions in one place” request is now bundled with that same work (read more).
zogstrip landed or referenced multiple fixes: a mobile scrolling issue in the reactions panel was already resolved with a linked commit (read more), continued investigating tag visibility edge-cases on Meta (read more), and tackled UI regressions like the backups breadcrumb separator race condition (read more) and sidebar icon search regression (read more).
supermathie weighed in on the “Python tooling vs automation” debate to clarify intent and acceptable experimentation boundaries (read more), raised an important identity/MFA propagation question for Discourse ID and federated login flows (read more), and filed a sharp bug report arguing outgoing emails should still link to topics even with “private email” enabled (read more).
david provided practical guidance on API tooling: public reads are supported (even without an API key), but bot requests face stricter rate limits and protections; the core message was to prefer keys where possible and keep usage respectful (read more).
Falco responded to embedded-comments feedback, highlighting mobile-first design intent and noting follow-ups still in progress (toolbar width, send-on-enter), and also linked a fix for an embed sizing bug via a core PR (read more). He also clarified “logged-out embed mode” behavior and pointed out Safari-specific consent prompts in a separate embedding support thread (read more).
manuel continued polishing the Meta-branded theme rebuild, noting icon work has been merged into the Lucide Icons component and playfully acknowledging “synchronous magic” when UI quirks seemed to resolve mid-report (read more).
pmusaraj explained that the standards-based path for passing MFA context is OIDC, while Discourse ID is currently OAuth2-only—laying out complications around IdP-provided MFA signals and the possibility of needing 2FA at the ID layer (read more). They also jumped into email-delivery troubleshooting, suggesting misconfiguration and asking about hosting context (read more).
sam gave the definitive protocol answer on OAuth/SSO URLs: a # fragment in a provider URL is a violation (fragments are client-side only), helping close out a puzzling configuration issue (read more).
awesomerobot shipped fixes for mobile moderation workflows after a detailed report, linking the PR that improves alignment/spacing and other reviewables issues (read more), and separately shared an “attempt at a fix” for modals opening above the viewport on mobile (read more).
“Our most exclusive support plan: just think about it, and it is fixed.” — manuel (read more)
And that’s a wrap—may your reactions menus scroll smoothly and your bugs fix themselves the moment you think about them.
Where the conversation clustered: the Meta homepage/layout refresh and early Blocks API rollout drove a lot of feedback (read more), while nested replies testing continued to surface UX/navigation edge cases (read more). Several admin/product updates also landed in #Announcements, including Data Explorer charts (read more) and more flexible approvals (read more).
“This was built using our new Blocks API… still an early preview.” — derek (source)
Interesting Topics
#Site feedback
The Meta Branded theme homepage got a dashboard-style refresh (events, hot topics, contributors, “watch & listen”), and feedback quickly honed in on who the new homepage is for and whether video content should be as prominent (read more).
Ongoing debate about Meta’s reorganized category structure continued: discoverability, long category paths in composer/autocomplete, and frustration linking #bug now that it lives under a parent category (read more).
A potential performance/UX snag was raised around disco-toc on very long, heading-heavy posts—safe mode appeared to render faster, suggesting a theme/component interaction worth digging into (read more).
In the “support categories” process discussion, Lilly argued that “mark as solution” often ends up incorrect or unhelpful in practice, with follow-on discussion about OPs marking their own posts and how to improve outcomes (read more).
#Announcements
The Data Explorer plugin gained more charting options and smarter auto-selection of chart types based on result “shape,” improving quick visualization of query outputs (read more).
Category moderation got a boost: topic/reply approvals can now be targeted (“everyone except group X” / “no one except group Y”) and settings were consolidated into a new Moderation tab in category settings (read more).
Bulk moderation got a practical upgrade: admins can now pin/unpin topics in bulk, choosing global vs category pin + timeframe in the bulk actions flow (read more).
Topic Voting improvements shipped: better vote button access while scrolling, opt-in notification prompts after voting, and “Hot” sorting refinements for ideas categories (read more).
A mobile UX papercut: a very long button label in the user deletion modal gets truncated; awesomerobot proposed a cleaner “select deletion type, then confirm” interaction pattern (read more).
Calendar/event display on mobile: an event’s date next to the title lacked styling and was confusing at a glance; a consistency fix was queued (read more).
“Empty space below unanswered topics” highlighted how reserved UI space (e.g., “person is replying”) can create layout inconsistencies between 0-reply and replied-to topics, prompting discussion of outlet placement and bounce prevention (read more).
A new UX suggestion: on narrow mobile viewports for anonymous users, make Sign up easier to find by placing login/sign-up actions in the hamburger/sidebar menu (linked to the long-running older request) (read more).
AI translation reliability concerns resurfaced: skipped locales (e.g., Portuguese) and how to handle transient LLM 503 errors without runaway token costs; staff clarified current retry/backfill behavior and expectations (read more).
Nested Replies testing kept revealing “find the unread replies” challenges and navigation expectations (e.g., a stronger “parent context” traversal), plus discussion of how nested reading shifts attention to branches, not “read all” (read more).
A neat formatting workaround: to “force” table columns to stop compressing (and become scrollable in practice), using (non‑breaking spaces) can prevent undesirable wraps in table headers/cells (read more).
derek introduced the updated Meta Branded theme homepage and engaged with feedback about content placement and logged-in vs logged-out behavior (homepage update; Blocks API note here).
awesomerobot emphasized that the homepage/layout changes aim to make Meta more welcoming beyond developers—particularly for admins and community managers (context)—and proposed a redesigned user deletion flow to avoid unwieldy button copy (mockups + approach).
lindsey shipped two admin workflow improvements: more flexible category-level approvals (details) and bulk pin/unpin for topics (how-to). She also announced a batch of Topic Voting UX improvements (read more).
nat posted the Data Explorer charting improvements announcement (read more), confirmed a localization fix for “one un-translated topic in an otherwise translated list” (fix deployed), and clarified AI tag translation scope/limitations (tags aren’t independently disable-able from other translation targets) (read more).
Falco continued to pressure-test nested replies UX, pointing back to the historic rationale for flat discussions while acknowledging gaps like better parent traversal in non-linear threads (nested replies thread; related context link here). He also followed up in the full comment system embedding area, noting Safari constraints and reiterating how Discourse’s automatic dark mode helps keep embeds in sync (embed follow-up; dark mode feature reference here).
markvanlan clarified that the flat/nested toggle is now behind a group site setting (staff-only by default) and discussed notification-linking challenges when multiple replies are consolidated (toggle note; notification issue analysis here).
supermathie updated several support threads: confirming a fix/merge for a Discourse MCP permission issue affecting about.json (status), recommending a safer GeoIP file management approach using symlinks rather than mounting the entire vendor data directory (suggestion), and advising that diagnosing a 500 error on an AI embeddings endpoint needs logs (Rails + /logs) rather than guesswork (read more).
pmusaraj responded on AI translation retry expectations and token-cost tradeoffs, noting Discourse avoids aggressive retries but does backfill/retry over time (read more; follow-up here). He also explained how to disable local logins while keeping OIDC signups workable (and how auth immediately affects the splash vs immediate redirect) (walkthrough).
sam jumped into a performance-heavy chat bug report with an optimization that produced a large improvement in testing, emphasizing the need to shed load safely even if counts become slightly approximate (read more).
jordan.vidrine suggested tightening confirmation UX copy: if the modal body already explains irreversible deletion, the button label can be shorter (e.g., “I understand”) (read more).
Yesterday’s Meta felt like a choose-your-own-adventure: new homepage blocks up top, nested reply branches down below—just try not to get lost in the thread forest.
Admins can now manage “permanently delete” without console support: lindsey shared an update that exposes can permanently delete to admins with extra safeguards, including explicit confirmation modals and a 5‑minute delay between soft delete and permanent delete actions (read more).
More charting options land for Data Explorer (and hosted rollout clarifications): follow-up discussion covered testing expectations and timing for hosted deployments, with staff confirming it should appear after the next deploy (read more).
#site-management
Auto-tagging topics using Discourse AI + Automation: a new guide explains how to configure AI triage to apply tags automatically, including prerequisites like enabling plugins and setting up an Agent persona (read more). Related docs include Discourse AI, Discourse Automation, the LLM settings page guide, and AI bot personas.
#plugin
Serve Discourse content as Markdown for LLM-friendly ingestion: benword released a plugin that returns “cooked” content as Markdown when clients request Accept: text/markdown or append .md to URLs, aiming to reduce token costs and simplify downstream processing (read more). pmusaraj highlighted the value of leveraging Discourse’s cooking pipeline for richer Markdown output (read more).
#theme-component
Theme component: “Show hidden categories” (with a nod to existing solutions): pfaffman shared a component that adds a “fake category” entry that links to a group page (to request access/join), and the thread quickly compared it to Category Previews and discussed possible differentiators (like category boxes support) (read more). Installation references: How to install a theme/component and the Beginner’s guide to using Discourse Themes.
Theme component: combined mobile “Sign Up / Log In” UX: Lilly introduced a component to replace the mobile login button with a combined flow, then followed up with an update that opens a modal with links to signup/login, crediting an idea from Moin (read more). This also connected back to the UX request around anon auth affordances on mobile (read more).
Google Photos album oneboxes look like single images: Ed_S reported that Google Photos album links onebox as a lone cover image with no clear indication it’s clickable or contains multiple photos; pmusaraj tagged it pr-welcome for contributors to improve (read more). The thread also referenced prior behavior/regressions in “Onebox support for new Google Photos”.
Content Localization SEO: ?tl= translations should be self-canonical: a report argued translated crawler pages are effectively “killed by canonical” because they canonicalize to the base language URL; nat agreed that each language variant should be self-canonical and said the team will investigate (read more).
“Each language variant should be self-canonical.” — nat (source)
Editing posts can explode raw size due to underscore alt-text escaping: a detailed report shows repeated edits double-escape underscores in image alt text, causing exponential growth in backslashes and eventually making posts uneditable due to length limits; zogstrip noted it’s likely a regression and started investigating (read more).
API DX feedback (headless frontends) + draft deletion auth inconsistency: one new thread proposed clearer pagination / post-stream separation for Next.js-style headless builds (read more), while another uncovered DELETE /drafts/{key}.json returning not_found via API-key auth unless ?username= is provided—prompting a consistency fix proposal (read more).
Reachable-mention warnings fail when usernames contain uppercase: thoka found that mentioning @SomeUser can bypass composer “not reachable” checks because the mentions endpoint returns lowercase usernames without user_reasons, so warnings never appear (read more). This investigation grew out of behavior observed in the long-running Private Topics Plugin thread.
AI translation + locale detection: Portuguese skipped and controls clarified: a deep dive described three overlapping issues (locale mis-detection, API failures, and category-targeting edge cases), and pmusaraj pointed out how to override locale via the composer controls and acknowledged category support gaps worth improving (read more).
#site-feedback
Meta homepage breakage fixed after a core change: Moin reported the homepage looking broken, and derek explained a recent core update changed how the custom homepage model is handled; the theme was updated and the issue resolved (read more).
#extras
Termcourse adds hi-res image expansion support (terminal-friendly upgrades): the terminal client gained better image expansion support (chafa/sixel/magick) for terminals that can display it, with implementation details linked via its PR (read more).
Webinar announced: “Community Retention: How to Define and Drive Return Behavior”: Danielle shared registration details and outlined a retention framework focused on return behavior (not just “more activity”), plus practical levers like onboarding moments and recognition (read more).
Sign-off: Yesterday’s highlights had everything—canonicals, oneboxes, and backslashes multiplying in the dark—just another day on Meta where the edge cases come out to play.
A second Domniq-adjacent showcase landed: a configurable premium landing page plugin for logged-out visitors (hero variants, community stats, trending topics, FAQs) in Domniq - A Premium Landing Page for Discourse Preview.
A small-but-impactful fix landed: “clearing advice” didn’t actually clear until a PR was merged; the merge confirmation is in Clearing advice doesn’t actually clear it.
#Feature
Incident responders asked for a notification level beyond Watching: a proposed “Obnoxious” level that could force desktop notifications on specific threads only (rather than enabling site-wide push) in New notification level: Obnoxious.
A request to make a useful moderation/power-user tool more shareable: the “view posts by @user” filtered-topic view should be linkable/shareable without insider knowledge in Make the “view posts by @user” filtered topic view shareable / linkable (with the current ?username_filters= behavior cited in-reply at post #2).
CI pain point: the minio_runner install script reportedly fails to follow a redirect, breaking shared GitHub Action workflows and leaving Rails system tests red in Minio installer breaks CI.
Helpdesk workflow scaling: a community running a busy support mailbox asked whether volunteers can “archive/hide” requests only for themselves (without archiving for everyone) in Support Mailbox archiving for some and not others.
#Theme
Theme breakage after core changes: the long-running “FKB Pro” theme thread reported a fatal frontend crash tied to the removed ember-this-fallback/js-component shim in FKB Pro - Social theme.
#Feature (Embedding / Analytics)
Embedding analytics surfaced sharp edges: a publisher reported double-firing analytics (page_view duplication, GTM scripts injected twice) when Discourse is embedded, with charts and detailed impact notes in Tecnoblog’s Experience With Discourse Comments.
In the last 24 hours, meta saw 69 new posts across 8 new topics. (Highlights include a new home for show-and-tell ideas and multiple product/UX updates: read more, read more, read more)
mcwumbly proposed a lightweight “post it in General and tag it show-and-tell” pattern for people sharing cool Discourse-related work—sparking immediate examples like a topic-embedding cluster map idea from piffy and a quick event countdown component from Lilly (read more, example post, example post).
#Announcements
Discourse is simplifying default email subject lines to feel less technical and reduce redundancy; early feedback included concerns about losing category context in group-based setups and when [PM] is helpful as an explicit indicator (read more, discussion, related: subcategory display options in email templates).
Emoji picker search now supports keywords in 48 languages, resolving a long-standing multilingual pain point and closing the loop with older requests about translated emoji aliases (read more, earlier request: translated emoji aliases, older support Q: other languages in emoji search).
A proposal to start a community-written “Meta Monthly Newsletter” drew pragmatic pushback: concerns about “user/staff of the month” turning into a popularity contest, and questions about what value it adds beyond official posts, AI summaries, and leaderboards (read more, critique](MNN Meta Monthly News - #4 by Canapin)).
A solved composer regression report (“composer ignores default_composition_mode”) turned into a plugin-hunt; safe mode was recommended to narrow down conflicts, with suspicion landing on third-party components (and a reminder that safe mode doesn’t require downtime) (safe mode guidance).
Embed users reported a serious “can’t type into the first comment” issue (plus iOS/Safari scroll quirks); follow-ups stressed impact on real readers and highlighted ongoing embed-related improvements being tracked by the team (read more, follow-up](Tecnoblog's Experience With Discourse Comments - #36 by Falco)).
When switching translation LLM settings mid-stream, a key question emerged: does Discourse redo translations? The answer: existing translations remain, and the new model handles only what hasn’t been translated yet (read more, answer](What happens to translations when LLM changes? - #3 by Falco)).
SEO + localization: confirmation that translated pages using ?tl= won’t rank as separate pages because canonicalization points search engines back to the original content (confirmation).
Admins changing a user’s email still hits awkward friction: verification links can fail with “Access Denied” while logged in as admin, prompting calls for a more seamless workflow (with workarounds via impersonation or console-based updates) (read more, console workaround reference](Change staff email address from console)).
#Feature
A thoughtful request explored “future self” / trusted-friends post approval for users with episodic unwellness; the conversation widened into ideas like “undo send,” scheduled posting, and even opt-in AI-based “are you sure?” drafting feedback (read more, undo-send tie-in](The awkward mishap of premature posting - #14 by NateDhaliwal)).
A bump request for emailing .ics calendar invites into group inboxes tied into ongoing Calendar/Events plugin improvements, connecting feature needs to broader project planning (read more, roadmap reference](Current Projects - April 2026)).
A user flagged a perceived problematic symbol in the Graceful theme’s background pattern; replies suggested it’s likely a stretch, and reminded that admins can swap the background if desired (read more).
Dev/install docs upkeep: a nudge that improvements to “Install Discourse on macOS for development” should likely be incorporated into the first post for discoverability (read more).
Self-hosting storage layout: clarification about disk usage split between /var/discourse/shared/standalone (forum data) and /var/lib/docker (images/overlay), as part of a question about putting them on separate volumes (read more).
lindsey announced upcoming default email subject line cleanups intended to reduce technical clutter and improve approachability, then invited feedback on edge cases like category/group context and PM labeling (announcement, community concerns](Simpler email subject lines - #2 by alehandrof)).
sam helped troubleshoot a “composer ignores default_composition_mode” report by encouraging safe-mode testing and isolating third-party plugin interference, leading to a likely culprit identified during debugging (diagnosis).
nat confirmed an important SEO implication for localized content: translated pages using ?tl= won’t rank independently due to canonical behavior (confirmation).
Falco weighed in across multiple fronts: explaining leaderboard anomalies can arise via private activity or external point-awarding systems (leaderboard context), reassuring that switching translation LLMs won’t trigger a full re-translation backlog (LLM translations), and shipping multilingual emoji search—then closing the loop by pointing older support threads to the new announcement (emoji announcement, back-reference](How to support other language in emoji search? - #2 by Falco)). They also followed up on embed/comment UX breakage reports in a high-impact publisher use case (embed follow-up).
That’s a wrap: yesterday’s meta felt like equal parts polish and play—tidying inboxes, teaching emoji search new languages, and still finding time to show-and-tell the fun stuff people are building.
kirupa showcased a show-and-tell experiment: a Discourse forum “revived” with AI bots that each have personalities and posting behaviors, sparking debate about what community means and whether bots can be a legitimate “daily briefing” tool (read more; discussion pushback: read more; team reaction: read more)
fuse wondered if Discourse should notify contributors when a topic is deleted or becomes inaccessible, after a thread seemingly vanished due to a privacy mishap—prompting a clarification and apology from staff (read more; staff follow-up: read more)
A quick shout-out landed for recent improvements to releases.discourse.org readability, with mcdanlj noting the site is “a lot easier to read now” (read more)
#Announcements
The emoji picker gained language-aware search, and follow-up questions immediately popped up: custom emoji naming/translatability and whether announcement tagging should include localization (read more; custom emoji question: read more; tagging suggestion: read more)
“Wiki rollback” turned out to be a permissions edge case: category moderators can’t use the “Revert Edit Revisions” button—even if they feel like “a moderator”—and the solution pointed straight at the permissions matrix (read more; solution: read more; permissions reference: read more; related wiki guide thread: read more)
AI translations troubleshooting: one admin reported huge daily token usage with zero translation progress, and Falco suggested enabling verbose logs and inspecting AI API audit-log entries to verify what’s actually happening (read more; staff guidance: read more; related ai plugin overview: read more)
An automation + ai “triage agent” kept flagging every post because the model appeared to treat “Examples” as part of the content under review; Falco recommended using Triage with AI Agent + the flag tool rather than forcing string-only responses (read more; staff workaround: read more)
A self-host install attempt on HestiaCP ran into port/proxy confusion plus a GitHub-related snag, with the OP sharing Nginx template details and asking which reverse-proxy approach is “right” (read more)
“Logs sent to wrong S3” got a pragmatic debugging angle: check for anything outside Discourse (bind mounts, cron, rclone, aws s3 sync) that might still be syncing to the old bucket even after config changes (read more)
Monetizing instructional videos: a new admin asked about selling paid access to videos inside Discourse; replies leaned toward hosting video offsite (e.g., members-only YouTube) or using a points/paywall-like plugin approach (read more; offsite hosting suggestion: read more; points-plugin idea: read more)
#Self-hosting
Disk layout flexibility: clarification that /var/discourse and Docker storage don’t need to live on the same volume—Discourse “doesn’t know or care”—as long as your ops story remains sane for future maintainers (read more)
#Community Building
A practical mini-guide showed how to connect DBeaver/TablePlus to Discourse Postgres by mapping a port, locking it down with firewall/Tailscale, and adjusting auth—aimed at safer “real GUI” DB access than the built-in Data Explorer (read more)
#Plugin
Locations plugin sponsors got an update: early-access invites went out, sponsor lists are reviewed daily, and a staged migration plan to public repo was reiterated (read more)
Trading Buttons plugin users reported a recent Discourse update seemingly disabled the plugin and removed category-level enablement options, raising compatibility and regression questions (read more)
#Theme
The #Theme thread for FKB Pro logged a post-upgrade break: after updating to v2026.4.0-latest / Ember v6.10.1, an error about missing ember-this-fallback/this-fallback-helper reportedly made the forum unusable—requesting a fix path (read more)
chapoi joined the bot-forum discussion, praising the styling and framing the bots less as “community” and more as an interesting way to stay updated on niche interests (read more)
mcwumbly responded to the “deleted/private topic” surprise, apologized for crossed wires, and indicated the team is untangling the situation so the missing content can be appreciated again (read more)
Falco handled two AI-adjacent support threads: advising verbose logging + auditing API log entries for AI translations that burn tokens without progress (read more), and recommending a more robust automation pattern (tool-calling via flag) for AI triage agents that were misusing “Examples” (read more)
Yesterday’s moral: whether you’re herding bots, reverting wikis, or chasing runaway translation tokens—meta was in full “debug the future” mode.
itsbhanusharma — 1 post / 1 like (noting parts of passwordless flow are already possible: read more)
merefield — 1 post / 1 like (showing off a “snazzy 3D view option” update: read more)
Falco — 1 post / 1 like (explaining why AI triage examples were mis-specified: read more)
@team standouts:mcwumbly (4 posts / 4 likes) and Falco (1 post / 1 like), per their activity in support + UX + community threads (read more, read more).
Interesting Topics
#Site feedback
nathank shared excitement about the direction of the Meta-branded theme rebuild—especially how the custom homepage work and emerging Blocks API tooling could help smaller orgs simplify their overall web stack (read more, read more).
#Feature
The “post approval by your future self / trusted friends” idea continued to evolve, with RGJ pushing back on framing this as simply “accountability training,” while ted broadened the use-cases (e.g., rage-bait, drunk-posting) but emphasized the need to work closely with the impacted user to validate what should be productized (read more, read more, read more).
Discourse ID forums got a practical update: mcwumbly confirmed users should now be able to change their username/name directly in user settings even if the account was created with discourse-id (read more, read more).
A new login feature request asked to decouple passwordless “email magic link” login/signup from enabling local password authentication, aiming for “local accounts + email verification + no local passwords” as a first-class configuration (read more). The replies ranged from “this is already possible” (read more) to the blunt suggestion that users should just use password managers (read more).
A related auth configuration question asked whether allowed email domains is validated against the email returned by OIDC at account creation time (openid-connect, login) (read more).
Discourse AI triage debugging: markschmucker reported examples causing “flags every post” behavior, and Falco explained the key nuance—examples are sent as prior turns, so they must mimic the expected LLM response (including tool calls), and for simple cases examples may be unnecessary if the prompt clearly describes when to call the tool (automationai) (read more, read more, read more).
On monetizing community video content, ted highlighted that once creators are paid, complexity increases beyond technical implementation into legal/operational considerations (read more, read more).
#Extras
A new tool, Discourse Shortlink Router, proposed a support-friendly way to swap long admin URLs for a short domain that redirects to the same path—followed by an intense privacy/trust back-and-forth about logging, cookies, and what the service operator can see (read more, read more, read more).
The ICS → Discourse Importer thread added a subtle but meaningful behavior note: the --time-only-dedupe option isn’t purely time-based because it still considers “close enough” locations—helping treat room changes as a “new topic surfaced in Latest” signal while reducing feed noise (calendar-and-eventrest-api) (read more, read more).
Jagster flagged that public edit history visibility doesn’t appear to cover changing the author/owner of a topic/first post—creating potential confusion for admins who post on behalf of users and then transfer ownership; mcwumbly suggested it looks like a bug or missing feature and moved it for UX discussion (read more, read more).
A mobile UI report about overlong titles not wrapping properly looks like it may already be fixed: mcwumbly tested across several themes (Horizon, Foundation, Meta Branded) and couldn’t reproduce, suggesting affected sites may just need an update (mobile) (read more, read more).
#Community Building
Planning for change management: a new thread asked how to reconfigure an existing community around a new product launch, with follow-up questions probing whether it’s a shift from single→multi product and what the motivator is—plus a suggested “audience venn diagram” approach (customer-success) (read more, read more, read more).
#Self-hosting / #Self-Hosting
An migrations-tooling thread reported IPBv4 migration pain points: avatar import inconsistencies, handling an already-existing admin user (and posts not migrating as expected), link rewriting gaps, reaction conversion inconsistencies, and even dropped text inside <a href> tags (read more, read more).
In the ongoing Cloudflare best-practices topic (how-tocloudflare), sandra.mccollum described two issues (a blank login screen on first access and embedded PDFs not rendering) that persist even when pausing Cloudflare—prompting a call for debugging strategies (read more, read more).
#Plugin
The long-running “Community Network Visualisation” plugin thread got a fresh highlight: merefield noted it’s received more love and now includes a 3D view option, complete with screenshots (read more, read more).
mcwumbly confirmed a recent improvement for discourse-id sites: users can now change username/name directly in settings even when created via Discourse ID (read more).
They also handled a UX-triage moment by flagging author-change visibility as likely a bug/missing feature and moving it for discussion (read more, read more).
On the community side, mcwumbly asked clarifying questions to better understand the context of a reconfiguration request for a new product launch (read more).
Finally, they suggested a likely resolution path for a mobile title-wrap report by noting it wasn’t reproducible on current themes and may already be fixed upstream (read more).
Falco diagnosed why Discourse AI triage “examples” were backfiring: because examples are sent as previous turns, they must match the expected LLM response format (including tool calls when relevant), and simpler prompts may work better without examples at all (read more, read more).
That’s a wrap—may today’s support links be shorter, your magic-link logins smoother, and your mobile titles finally wrap the way they should.