Weekly Summary

News and Events Forum summaries
120

This week on meta.discourse.org…

Key Stats

Reporting window: 2026-04-19 → 2026-04-26 (last 7 days)

  • New posts: 615
  • New topics: 82

Top users (posts + likes received)

User Posts Likes
Lilly 41 110
Canapin 33 78
lindsey 7 75
Moin 37 68
Falco 22 66
nat 11 53
awesomerobot 11 40
Ed_S 17 40
merefield 15 40
chapoi 10 36

A few threads that drove a lot of discussion this week (sampling)

  • The Meta theme homepage got a Blocks-powered refresh, plus follow-up troubleshooting around “why don’t I see it when logged in?” (read more)

  • Reactions/likes UX changes kept evolving, including fixes and follow-ups for mobile/touch behavior (read more)

  • Nested Replies testing continued to surface real-world navigation + notification challenges (and feature debate) (read more)

  • Several admin-focused improvements landed in quick succession: Data Explorer charting (read more), category approval flexibility (read more), Topic Voting UX improvements (read more), and bulk pin/unpin actions (read more)

  • Internationalization and SEO got tangible wins: localized emoji search (read more) and a canonical/hreflang fix for ?tl= pages (read more)

Interesting Topics

#announcements

  • Current Projects (April 2026) landed with the latest “what we’re building now” roundup, framing recent deliverables like embedding improvements and ongoing focus areas for the platform (read more)

  • Data Explorer got more charting options: nat shared how the plugin now auto-selects better chart types based on result “shape”, making queries easier to interpret without exporting data elsewhere (data-explorer) (read more)

  • Category approvals became more flexible: new category moderation controls allow “everyone except certain groups” or “only certain groups” approval requirements, reducing mod workload while keeping guardrails (#announcements) (read more)

  • Topic Voting UX improvements shipped: voting access while scrolling, prompts to follow after voting, and improved “Hot” sorting made idea categories feel more responsive and intentional (topic-voting) (read more)

  • Bulk pin/unpin is here: admins can now pin or unpin multiple topics at once from topic list bulk actions—small change, big time saver for ongoing curation (#announcements) (read more)

  • Simpler email subject lines: lindsey outlined changes that remove redundant/technical formatting from email subjects to make communities feel less “developer-ish” in inboxes (email) (read more)

  • Emoji search now works in your language: emoji keyword matching expanded beyond English into 48 locales—a long-standing need for multilingual communities (localization) (read more)

#site-feedback

  • Meta’s homepage got a dashboard-style update: derek rolled out a Blocks-based homepage layout (events, hot topics, contributors), with discussion on why educational video content is surfaced and how personal homepage preferences interact with the new “Home” link (meta-redesign-2026) (read more)

  • Where should “show-and-tell” posts go? mcwumbly suggested a lightweight convention: post in General and tag show-and-tell—then let patterns emerge before adding structure (and the topic itself became a mini show-and-tell) (read more)

  • Category organization feedback continued: users highlighted friction around long inline category references (e.g. parent/subcategory paths), hashtag completion ergonomics, and discoverability tradeoffs after the Meta category refresh (meta-redesign-2026) (read more)

  • Restricted/private-ish tags leaking into UX: the “tag cleanup” thread resurfaced questions about whether restricted tags should be discoverable to unauthorized users at all—leading to investigation and a core fix (tags) (read more)

  • Possible slow rendering with disco-toc + many headings: Moin reported noticeable delays that disappeared in safe mode, raising suspicion that certain components (or ToC generation) may be too heavy on long/structured posts (disco-toc) (read more)

ux (and UX-adjacent buglets)

  • User deletion modal had an absurdly long button label on mobile: Canapin reported the truncation, and awesomerobot proposed a redesigned flow (choose delete type first, then confirm) with a PR already in motion (mobile) (read more)

  • Event date styling next to topic titles was confusing on mobile: the mobile rendering lacked the visual separation seen on desktop, prompting a quick “make it consistent (and simpler)” fix proposal (calendar-and-event) (read more)

  • Mobile review queue UI got a cleanup pass: misaligned header elements, spacing issues, and overly tall titles were raised—then awesomerobot followed up with a batch of improvements in core (review-queue) (read more)

bug / Support

  • SEO + localization bug: ?tl= pages canonicalized incorrectly: multilingual admins flagged that translated crawler pages couldn’t rank because the canonical pointed to the base language; nat agreed and shipped a fix via core PR (seo, content-localization) (read more)

  • Chat performance: Pitchfork timeouts and “silent threads”: a deep write-up suggested chat reply tracking could bloat over time and cause slow/unreliable loads; sam tested and reported a huge performance improvement via indexing/query changes (chat) (read more)

  • Mentions reachability check broke with uppercase usernames: thoka found case-sensitivity gaps that prevent “won’t notify” warnings in restricted categories; follow-up discussion led toward smaller, safer endpoint-level fixes (mentions) (read more)

#plugin / #theme-component

  • Communiteq Power Tools plugin: RGJ consolidated lots of “small but practical” admin enhancements into one maintained plugin—reducing the need for console access and bundling quality-of-life tweaks (#plugin) (read more)

  • Serve Discourse content as Markdown: benword released a plugin that outputs “cooked → markdown” via Accept: text/markdown or .md URLs, aimed at cheaper LLM ingestion and simpler downstream processing (markdown, ai) (read more)

  • Combined login/signup button component: Lilly shipped a header-space saving theme component that merges auth buttons into one dropdown/modal—especially helpful on mobile where signup can be hard to spot (login, signup) (read more)

  • Show hidden categories component: pfaffman shared a theme component approach to “preview” hidden categories (via a “fake category” linking to group join/request), and compared it to existing options (#theme-component) (read more)

General / #extras (show-and-tell energy)

  • Domniq: a Discourse-native Android+iOS app preview: nicolsdennis showcased an app build with push notifications, moderation actions, and a polished UI—plus discussion of theme inspiration and chat styling (show-and-tell) (read more)

  • “I populated my forum with bots” sparked a philosophical debate: the experiment reframed bots as a personal briefing engine rather than simulated community, and the thread explored authenticity, value, and design aesthetics (show-and-tell) (read more)

  • Discourse Shortlink Router proposed a new way to share “support URLs” cleanly: swap your domain for a router domain to generate clickable admin/settings links, alongside a trust/privacy discussion (#extras) (read more)

Activity by the @team Group

  • lindsey shipped multiple admin-facing improvements and updates this week: the April roadmap newsletter (Current Projects), more flexible category-level approvals (read more), bulk pin/unpin actions (read more), and Topic Voting enhancements (read more). She also continued the “make Discourse feel less technical in inboxes” push with simplified email subjects (read more) and documented new safeguards for permanent deletion settings becoming admin-visible (read more).

  • Falco advanced multilingual UX with localized emoji search (read more) and pointed translators back to the change when older support topics resurfaced (read more). He also kept driving embed-as-comments improvements and troubleshooting in the Tecnoblog embed thread (read more), helped isolate tracking/analytics issues via PR links (read more), and advised on Ghost integration behavior and expectations (read more). On AI translation operations, he clarified how switching LLMs affects translation progress (read more) and coached on improving AI triage examples/tool-call usage (read more).

  • nat focused on localization correctness and SEO: she confirmed that translated ?tl= pages should be self-canonical and shipped a fix quickly (read more), plus guided follow-up rollout questions (read more). She also clarified retry/backfill behavior and expectations in an AI translation reliability debate (read more), and closed the loop on multiple content-localization bugs with deployed fixes (read more; read more; read more). Separately, she announced richer Data Explorer charting to make community analytics more accessible (read more).

  • awesomerobot had a particularly UX-heavy week: he proposed a cleaner deletion flow to avoid giant mobile button labels (read more), addressed mobile event date styling inconsistency (read more), and pushed fixes for mobile review-queue alignment/spacing issues (read more). He also took on tricky mobile modal/viewport behavior with an “add user to list” modal fix attempt (read more) and explained why some “empty space” on topic lists is reserved for live presence UI (read more). On Meta’s redesign goals, he added important context about broadening Meta beyond developers in the homepage refresh thread (read more).

  • chapoi continued polishing the reactions/likes experience: he shared progress on making “display all reactions” better than the initial fix (read more) and clarified intended “new normal” behavior (reactions on the left) (read more). When users of likes-only sites objected to the new two-heart UI, he confirmed the team was working on restoring the single-like-button layout for reactions-disabled sites (read more). He also chimed in on Meta tag UX expectations around restricted tags (read more) and reacted to the AI-bot “show-and-tell” forum experiment with thoughtful framing (read more).

  • derek rolled out and supported the new Blocks-based Meta homepage, explaining design intent and tradeoffs (including surfacing educational content) (read more), and then quickly resolved a break caused by a core change interacting with the theme’s custom homepage model (read more).

  • zogstrip was active across multiple “small but sharp” bugfixes: he investigated restricted tags leaking into tag autocomplete and landed a core fix (read more), reviewed the “double-escaping underscores in image alt text” regression and opened a PR despite it being tricky (read more; read more), and handled draft deletion API inconsistencies with a fix proposal (read more). He also responded to mention-case-sensitivity issues by opting for safer incremental endpoint fixes (read more) and investigated a backups-page breadcrumb race condition (read more).

  • mcwumbly helped shape Meta’s culture/structure around “sharing builds” by proposing the show-and-tell convention (read more) and then encouraging spin-off topics for deeper feedback loops (read more). He also explored product-design space in a sensitive feature request about “post approval by future self / trusted friends,” including the idea of opt-in just-in-time AI feedback (read more), and apologized/clarified a Meta mishap where a topic became inaccessible due to category privacy changes (read more). Finally, he confirmed Discourse ID username changes are now available directly in user settings after recent work (read more).

  • supermathie tackled several technical/support edges: he highlighted how “private email” mode makes outgoing emails confusing when the topic link is removed (read more), participated in an older-but-active discussion about automating access to Discourse via external tools and clarified intent vs automation boundaries (read more), and dug into federated 2FA/MFA signaling limitations for Discourse ID (OAuth2 vs OIDC needs) (read more). He also confirmed an MCP permission fix for about.json access was merged (read more) and answered a self-hosting filesystem question with a pragmatic GeoIP symlink suggestion (read more).

  • pmusaraj stayed deep in Discourse AI and “diagnose + set expectations” support work: he explained token-cost implications and retry behavior in translation failures when upstream LLMs return errors (read more), and followed up with concrete improvements to category-targeting behavior and locale overrides in AI translation tooling (read more). He also encouraged contributions on Onebox behavior for Google Photos albums by tagging it pr-welcome (read more), and praised the “cooked-to-markdown” approach in the new Markdown export plugin thread (read more).

Thanks for reading, and I’ll see you again next week! :slight_smile: