Patch Notes

Improved the accuracy of how GameDevPress interprets reported game-industry sales and revenue milestones, including forecasts, denials, and corrections.

Impact

Forecasts, ambiguous reports, denials, and corrections no longer receive the same milestone authority as a completed, unambiguous result; independent studio, funding, labor, acquisition, and policy stories remain eligible under normal feed rules.

User action

Nothing.

Archive page snapshots now stay available while GameDevPress builds a validated replacement in the background.

Impact

Readers can keep browsing a stable archive page during cache refreshes instead of encountering a temporary archive-page outage.

User action

Nothing.

Improved source filtering so material game-development news is less likely to be excluded before it reaches the feed's regular ranking.

Impact

Readers should see more meaningful studio, platform, development, launch, and major-update news considered by the existing feed pipeline, while guides, reviews, giveaways, discounts, routine patch notes, and low-context trailers remain filtered.

User action

Nothing.

Archive navigation now shows your current page and total pages, with compact controls and a direct page jump for reaching older GameDevPress stories.

Impact

Readers can move directly through the long-lived archive with shareable page links while live pagination keeps a stable snapshot during a browsing session.

User action

Use the page controls or enter a page number to jump to older stories.

GameDevPress can now preserve completed story, article-metadata, source-provenance, tag/entity, and compact ranking-audit records in a durable private archive with manifest-backed verification.

Impact

Operators can retain long-term research evidence while the live service stays focused on current D1-backed feeds. No public bulk-download access was added.

User action

Nothing.

Added App2Top as a watched trade/regional source for material game-industry business, market, legal, funding, and studio coverage, with filters to keep iGaming and promotional posts out of the feed.

Impact

Readers have another path to see relevant regional and business coverage while routine milestone posts, iGaming coverage, and investor/publisher promotions stay filtered.

User action

Nothing.

Added an admin-only dry-run D1 telemetry retention report that uses storage diagnostic snapshots to count old private selection-audit JSON candidates, protected exclusions, blocked reason buckets, and estimated eligible bytes.

Impact

Operators can plan retention for bulky private audit data before it threatens ingestion while public feed, readiness, archive, tag, ranking, and frontend behavior remain unchanged.

User action

Nothing.

Reduced repeated story cards across latest, tag, and archive feeds so the same article or story chain appears once while genuine material follow-ups can still surface.

Impact

Readers should see cleaner feed, tag, archive, and Daily Review story streams without losing meaningful updates that add new facts.

User action

Nothing.

Added admin-only D1 storage growth diagnostics with compact snapshots, top table and JSON/text growth drivers, trend state, and configurable budget warnings.

Impact

Operators can spot storage pressure before it threatens ingestion while public feed, readiness, archive, ranking, and selection behavior remain unchanged.

User action

Nothing.

Daily Review can now show a published Hermes-authored Daily Summary above Key Themes when editorial text exists for the completed UTC date.

Impact

Readers get a concise Hermes-authored editorial overview without changing story grouping, themes, videos, or the sanitized public API behavior when no summary has been published.

User action

Nothing.

Added The Verge as a carefully filtered source candidate for platform policy, distribution, hardware strategy, regulatory, market, and game-industry business stories.

Impact

Readers can see more high-impact Verge coverage when it affects game development or the games business, while routine consumer tech reviews, deals, how-to, entertainment, and podcast posts stay suppressed.

User action

Nothing.

Deterministic selected-story tag enrichment now records private evidence and diagnostics without finalizing public tags or clearing trusted Hermes decisions on its own.

Impact

Operators can distinguish stories still waiting on Hermes from stories Hermes intentionally left tagless, while public tag payloads stay sanitized.

User action

Nothing.

Every newly selected story now gets a durable Hermes tag-curation target with admin-visible pending, applied, empty, failed, blocked, and stale lifecycle state.

Impact

Operators can track missing or delayed story-tag curation without blocking public feed freshness, and tag changes still require the governed Hermes/admin apply path.

User action

Nothing.

Selected-story tag diagnostics now distinguish missing Hermes decisions from explicit empty decisions, successful applies, failed/projection-failed applies, and stale target evidence.

Impact

Operators can spot tag coverage gaps instead of treating every empty public tag list as successful curation, while public APIs remain sanitized.

User action

Nothing.

Hermes story-tag curation now keeps active authority-resolved desired tags instead of dropping them when backend phrase or semantic-support checks disagree.

Impact

Operators can trust Hermes desired display, index, and mention-only tag decisions to reach the selected-story tag read models, while stale targets, authority conflicts, privacy, projection, and rollback failures still block safely.

User action

Nothing.

Added cautious filtered RSS coverage for MassivelyOP and Gameoneer to catch more MMO, live-service, and niche-market development signals.

Impact

The feed can pick up material shutdown, studio, publishing, monetization, policy, and launch-market stories from these outlets while routine guides, trailers, reviews, patch notes, and community posts stay suppressed.

User action

Nothing.

Selected-story public tags and tag timelines now come only from Hermes/governed story-subject decisions, while automatic enrichment keeps evidence and candidates private for review.

Impact

Readers should see fewer guessed public story tags, and operators still get private diagnostics and suggestions for trusted tag curation.

User action

Nothing.

Selected-story tagging now gives exact headline game and hardware-product subjects stronger authority before broader context tags are finalized.

Impact

Readers are less likely to see specific story subjects hidden behind broad company or lifecycle tags when the headline clearly names the game or device.

User action

Nothing.

Hermes/admin story-tag curation can now repair selected-slot tag JSON and exact public tag read models when durable story subjects already match the desired state.

Impact

Operators get report-only and dry-run evidence for bounded projection repairs, and apply runs can clear stale public tag projections without hand-editing archive data.

User action

Nothing.

Quiet-feed drought readiness now requires the Worker status itself to be healthy before an in-bound no-candidate stretch can be accepted.

Impact

Operators get a failed health signal when a quiet-feed payload is paired with a failed Worker run, while legitimate bounded quiet periods still pass without weakening story selection.

User action

Nothing.

Hermes/admin story-tag curation now plans report-only, dry-run, and apply decisions through the same guarded path.

Impact

Operators can apply safe story-tag fixes even when unrelated current or cleanup tags are omitted, while stale targets, stale or adjacent support, authority conflicts, projection failures, and other integrity failures still block safely.

User action

Nothing.

Added a public Feed Events tracker near Patch Notes so readers and operators can see recent feed checks, published stories, and safe reasons when no new story cleared the quality bar.

Impact

Readers can distinguish a healthy quiet feed from delayed, degraded, or failed checks without exposing private operational diagnostics.

User action

Nothing.

Explicit acquisition, minority-stake, and material studio-investment stories can now show an Acquisition tag alongside the involved company tags.

Impact

Readers get clearer story-card context for current acquisition coverage while user-acquisition, gameplay-item acquisition, background mentions, and generic investment stories stay filtered out.

User action

Nothing.

Hermes/admin story-tag curation now recognizes safe same-story aliases, plural forms, material support fields, and bounded company-name variants before rejecting direct evidence as unsupported.

Impact

Operators can land clearer direct story subjects such as products, studios, closures, remasters, and material projects while wrong-family aliases and broad authority growth remain governed.

User action

Nothing.

Hermes/admin authority requests now report governed create, promote, propose, and block recommendations while keeping broad IP or franchise-family gaps proposal-only.

Impact

Operators get clearer private diagnostics for missing game, product, initiative, and related authority without changing public feed, tag, ranking, or readiness payloads.

User action

Nothing.

Production and admin recovery smoke guidance now requires a fresh-main verification context before operators interpret broad live smoke results as API or admin health.

Impact

Operators get an early stale-context failure instead of mistaking an outdated Daily Review validator for a production regression or admin-token problem.

User action

Nothing.

Improved the feed refresh path so recent articles can be selected through an indexed recency key as the article archive grows.

Impact

GameDevPress can keep refresh latency more predictable without changing feed ranking, source rules, public APIs, or the visible site experience.

User action

Nothing.

Daily Review cards are cleaner and easier to scan: the inline Why picked explanation was removed while story tags and related coverage remain available.

Impact

Readers can move through Daily Review story cards faster while still seeing tags, outlet context, and related articles.

User action

Nothing.

Daily Review now highlights up to eight clickable Key Themes from the public tags on the selected written stories for that day.

Impact

Readers can scan the day's dominant topics and jump straight to related tag coverage without changing story selection, ranking, or tag generation behavior.

User action

Nothing.

Story tags are now clickable, opening a shareable timeline of selected stories that used the same public tag across Latest, Archive, and Daily Review.

Impact

Readers can jump from a visible tag on any story card to that tag's selected-story history without manually searching the archive.

User action

Nothing.

Tightened story-tag repair checks so public tags stay attached to the exact selected story, source/category boilerplate is blocked from becoming game authority, and the affected June 30 stories can be repaired through governed curation.

Impact

Readers get more trustworthy story tags and operators get safer Hermes/admin correction evidence. No user action is required.

User action

Nothing.

Backfilled the visible-tag timeline read model for historical selected stories using only tags already shown publicly on story cards.

Impact

Tag clicks can now return the complete selected-story history for those public visible tags instead of missing older tagged stories.

User action

Nothing.

The homepage feed status now explains when the live Worker is still checking sources but no publishable story has cleared the quality bar beyond the normal quiet-feed bound.

Impact

Readers and operators see the actual public-safe reason for a quiet feed instead of a generic uncertain-status warning.

User action

Nothing.

Hermes story-tag curation now blocks active wrong-family tag or alias authority before applying selected-story tag changes.

Impact

Operators get explicit conflict telemetry when a requested company, game, product, topic, or region would resolve to the wrong tag family, while same-family aliases can still apply with requested and resolved identities recorded privately.

User action

Nothing.

Hermes tag curation now projects safe warning-level display and index tags instead of silently omitting them when authority was missing.

Impact

Readers can see intended game and company context after trusted curation, while operators still get private warning telemetry and hard conflicts continue to block safely.

User action

Nothing.

Hardened Hermes tag curation projection so multi-lane desired tags persist consistently and failed applies restore prior selected-story tags instead of clearing them.

Impact

Operators can apply trusted curation with more than eight display, index, and mention-only evidence rows while projection failures roll back selected-slot tags and tag read models to their prior state.

User action

Nothing.

Hermes tag curation now treats editorial authority uncertainty as warning telemetry, so safe desired tag updates can land.

Impact

Operators can apply trusted Hermes display and index curation even when some tags are broad, topic-like, short, low-confidence, new, or omitted for non-conflicting authority reasons, while duplicate, privacy, target, projection, and rollback integrity failures still block.

User action

Nothing.

Hermes tag curation now applies safe display and index tag updates even when exploratory mention-only authority is still unresolved.

Impact

Operators get applied public tag fixes plus structured mention-only authority-gap telemetry instead of losing useful display tags to non-conflicting mention-only gaps, while conflicts and projection failures still block safely.

User action

Nothing.

Curator apply now preflights required tag authority changes before writing story tags or creating authority rows.

Impact

Blocked Hermes curation attempts leave story tags and authority rows unchanged unless an unresolved authority was explicitly optional and safe to omit, making operator retries safer.

User action

Nothing.

Added private Hermes desired-state tag reconciliation so operators can apply complete story tag updates, including add, remove, demote, promote, and replace decisions.

Impact

Selected-story tag corrections now update private subjects, public tag rows, tag timelines, and selected-slot payloads together after exact target and authority checks, while public APIs stay sanitized and private before/after evidence is recorded.

User action

Nothing.

Added private append-only story-level Hermes curator telemetry with bounded admin diagnostics for recent decisions and failure buckets.

Impact

Operators can inspect sanitized curation attempts, replay/conflict state, contribution links, and safe reason buckets while public feeds, tag payloads, readiness, ranking, and ingestion remain unchanged.

User action

Nothing.

Added a bounded authority-request policy for Hermes canonical tag creates with exact no-op replay, required rationale/evidence, safe family checks, alias caps, and held/blocked conflict outcomes.

Impact

Operators can safely add clear missing authority such as game or company tags only when bounded rationale and sanitized evidence are supplied, while unsafe prefix, alias, topic, franchise, cap, missing-evidence, and privacy cases stay out of public read models.

User action

Nothing.

Added a private story-level Hermes tag curation validation endpoint for exact selected stories.

Impact

Operators can validate governed multi-tag curation payloads while public feeds and public tag payloads remain unchanged.

User action

Nothing.

Improved Hermes story-context safety so exact latest-feed and archive lookups keep the requested selected-story coordinates instead of falling back to another story.

Impact

Operators can run safer report-only Hermes scans across non-top feed stories, with mismatched context still blocked before any future story-tag apply path.

User action

Nothing.

Improved feed guardrails so Prime Day, percent-off, cheap, coupon, retailer, toy, merch, hardware, and accessory deal posts are treated as consumer shopping noise instead of material game-dev business stories.

Impact

Readers should see fewer affiliate-style offer pages crowding out industry news, while publishing, M&A, funding, labor, legal, platform-policy, market-data, and sales-analysis stories remain eligible.

User action

Nothing.

Added a public-safe homepage feed status line that distinguishes recently selected stories from active monitoring periods where no high-signal story was worth publishing.

Impact

Readers can tell when GameDevPress is still checking for important stories without seeing private operations details or lowering the feed quality bar.

User action

Nothing.

Hermes Tag Ops can now submit governed canonical tag authority requests through its admin tool wrapper without targeting a selected story.

Impact

Operators can grow safe tag authority separately from exact story-tag corrections, while selected-story JSON, public tags, timeline read models, ranking, archive, and readiness remain unchanged by authority-only requests.

User action

Nothing.

Improved feed ranking for layoffs and restructuring stories so material labor and studio-structure news is easier to catch.

Impact

The boost is bounded and requires subject-grade tags or title/summary evidence with a concrete affected subject, while duplicate, repost, low-confidence, and routine coverage remains constrained.

User action

Nothing.

Moved Game World Observer, GamesBeat, and MobileGamer.biz back to regular news classification while keeping specialist analysis sources curated.

Impact

Readers should only see the curated analysis card treatment on the dedicated analysis and specialist insight sources, not broad news, business, or mobile coverage from those three outlets.

User action

Nothing.

Service health now treats a fresh no-candidate feed as a bounded quiet-feed state for up to 18 target slots instead of paging solely because the latest selected story is older than the normal freshness window.

Impact

Operators can distinguish a healthy market lull from stale Worker status, source starvation, failed finalization, new-article no-selection problems, or quiet feeds that exceed the operational bound, while selected-story age remains honest and publication standards stay unchanged.

User action

Nothing.

Disabled BlockchainGamerBiz as an active scanned source so new feed updates focus on the current trusted source mix while preserving existing archive and source history.

Impact

Readers should keep seeing the feed refresh from the remaining active sources without new BlockchainGamerBiz candidates, while operators retain historical archive and provenance context for older coverage.

User action

Nothing.

Curated analysis cards now get a restrained feed treatment and visible label so specialist insight is easier to recognize while scanning selected stories.

Impact

Readers can spot curated industry-analysis picks without changing story order, selected-story eligibility, summaries, tags, coverage links, or the hot-video card treatment.

User action

Nothing.

Story cards now show a sparse Hot badge when an already-selected homepage or archive story has unusually strong impact signals.

Impact

Readers can spot the biggest game-dev stories faster while routine catalog updates, low-value filler, weak consumer noise, thin trailers, stale items, and duplicate variants stay unbadged.

User action

Nothing.

Added a governed admin path for Hermes to submit canonical tag authority requests without applying story tags.

Impact

The path reports exact matches and unsafe nearby matches, records rationale and evidence for audit, and holds ambiguous names for operator review while public feed tags and read models stay unchanged.

User action

Nothing.

Scheduled service-health failures now write sanitized JSON and Markdown reports that split infrastructure, ingestion, selected-feed freshness, and optional-lane context, then deduplicate repeated GitHub alerts by stable failure fingerprints.

Impact

Operators can tell which health lane failed, see the freshness budget and bounded admin terminal-state context when available, and avoid noisy duplicate incident issues without exposing secrets or private diagnostics.

User action

Nothing.

Hermes tag corrections now preserve useful existing story-card tags during additive tag actions and require governed rationale for removing, demoting, promoting, or replacing existing tags.

Impact

Readers keep material context such as lawsuit topics when safe company tags are added, while operators get audited correction paths for noisy or wrong tags without changing the public tag payload shape.

User action

Nothing.

Pull requests now get non-deploying Worker/Site validation contexts for Python tests, Worker typecheck and dry-run, focused freshness invariants, and the static site build.

Impact

Operators can see missing, pending, or failed PR checks in release evidence before merge without exposing secrets, running admin actions, deploying Cloudflare changes, or changing feed selection behavior.

User action

Nothing.

Production admin smoke now validates private selection-outcome evidence for recent final slots and the current final slot when admin readiness exposes one.

Impact

Operators get a stronger release gate for feed machinery without exposing private diagnostics through public readiness, status, latest, archive, or Pages payloads.

User action

Nothing.

Operator tag-authority lookup now returns exact canonical tags and active aliases, including Nintendo Switch 2, Doom, and Artificial Intelligence, before fuzzy or substring matches.

Impact

Hermes report-only readiness is safer for known authority cases while live tag mutation still requires an explicit canary and public tag payloads remain unchanged.

User action

Nothing.

Admin readiness now includes a bounded freshness telemetry summary that classifies feed machinery as healthy, quiet, source-starved, delayed, stuck, failed, or unknown.

Impact

Operators can distinguish quiet no-candidate periods from source, queue, finalizer, pending-work, and failure conditions behind admin auth while public readiness, status, latest, and archive payloads stay sanitized.

User action

Nothing.

Refresh slots and private selection outcomes now record bounded machine-readable reason codes for empty, failed, stale, skipped, and recovered feed refresh paths.

Impact

Operators can diagnose repeated no-candidate, source failure, queue, deadline, stale, and overlap outcomes from D1/admin diagnostics without relying on transient Worker logs, while public readiness and feed APIs stay sanitized.

User action

Nothing.

GameDevPress now uses Jason Schreier's Bluesky profile as the high-signal discovery source for his games-industry reporting instead of scanning Bloomberg's paygated gaming feed by default.

Impact

Readers should have a better chance of seeing timely Jason Schreier reporting when it is relevant to game developers and industry watchers, while personal, reposted, off-topic, and paywall-confusing items stay out of the selected feed.

User action

Nothing.

When normal feed selection finds no publishable story for longer than the documented quiet-feed freshness bound, the Worker can now recover with a recent high-signal industry/business story that passes strict anti-duplicate and noise guardrails.

Impact

Readers should be less likely to see a stale feed during extended quiet periods, while duplicate repeats, stale durable articles, utility posts, thin trailers, routine catalog churn, and weak reaction stories remain suppressed.

User action

Nothing.

Improved Hermes tag authority lookup for known active subjects and added a bounded admin HTTP canary that verifies exact story targets and public tag projection before mutation readiness is claimed.

Impact

Operators can test add, remove, replace, and safe tag-creation paths through the governed admin endpoint while live mutation cron remains disabled and public tag payloads stay limited to slug, label, and type.

User action

Nothing.

High-signal interview stories with meaningful industry, executive, strategy, market, or original-reporting context can now rank more strongly in the feed.

Impact

Readers should be more likely to see substantive interview-led game-dev stories, while generic promotional Q&A, thin transcripts, profiles, and routine marketing interviews remain guarded by the normal feed-quality filters.

User action

Nothing.

Feed selection now checks the latest public story window as well as the wall-clock cooldown so exact story, duplicate-chain, title/source, and canonical URL repeats stay suppressed while the earlier card is still visible.

Impact

Readers should see fewer repeated stories during quiet periods when the latest 24 cards span more than 24 hours, while genuinely new material follow-ups can still appear.

User action

Nothing.

Improved the private tag authority lookup used by Hermes so canonical tags, curated aliases, active aliases, slugified labels, and read-model-only gaps resolve with deterministic ranking and bounded diagnostics.

Impact

Operators get safer pre-mutation checks for ambiguous labels such as AI and alias targets such as Nintendo Switch 2 without enabling live mutation cron or changing public tag payloads.

User action

Nothing.

Improved feed selection to suppress consumer-facing upgrade, entitlement, guide, service-notice, and shopping/deal posts unless they include real game-dev, platform, business, labor, legal, market, or strategic industry signal.

Impact

Readers should see fewer player utility, free-upgrade, claiming, affiliate-shopping, and hardware-deal posts displacing industry-relevant game-dev news.

User action

Nothing.

Added governed admin archive-story inspection plus exact selected-slot archive tag correction through the existing audited Hermes tag contribution path.

Impact

Operators can safely inspect and correct archived selected-story tags with dry-run defaults, exact archive-date validation, contribution history, and unchanged public tag payload fields.

User action

Nothing.

Expanded private Hermes report-only tag authority with governed AI, revenue, market-data, and obituary topics plus safer aliases for layoffs, lawsuits, and Switch 2.

Impact

Operators get clearer report-only Hermes decisions for common material story gaps without enabling live mutation or changing public tag payload fields.

User action

Nothing.

Added private admin context and report-only safeguards for future unattended Hermes tag operations, including preserved existing story tags and governed material-topic authority.

Impact

Operators can inspect bounded selected-story context and dry-run trusted tag corrections without changing public tags, authority, aliases, selected-slot JSON, readiness, or publication state.

User action

Nothing.

Trusted Hermes tag contributions can now create safe missing game, product, company, platform, or event authority and apply it to the exact current selected story when deterministic safety checks pass.

Impact

Operators can grow the live story tag dictionary through the governed admin path without changing the public tag payload shape, while people, weak evidence, conflicts, stale targets, and privacy-sensitive requests stay blocked or proposal-only.

User action

Nothing.

Added the admin tag-contribution foundation for trusted live story-card tag corrections, governed tag and alias authority growth, and private correction history for future Hermes-assisted Tag Ops.

Impact

Operators can safely record, propose, apply, undo, and audit exact selected-story tag corrections while public story-card tag payloads remain limited to slug, label, and type.

User action

Nothing.

Added a local operator scorecard that checks whether subject-first tags meet the rollout gate before downstream tag surfaces can move forward.

Impact

Operators can produce sanitized PASS, FAIL, or INSUFFICIENT_COVERAGE summaries with subject recall, precision, noise, feed-fit, coverage, and privacy checks while public tag pages, search, ranking, related stories, rollups, and payload expansion remain blocked by default.

User action

Nothing.

Selected-story enrichment can now keep private scoped memory of repeated subject-tag noise and use it to suppress recurring background, related-link, evergreen, parent-company, broad-topic, and not-story-subject display mistakes.

Impact

Readers should see fewer misleading story-card tags while true headline and lead subjects still display. Public tag payloads remain unchanged and expose only slug, label, and type.

User action

Nothing.

Story-card tags now come from durable reviewed story subjects instead of transient extractor output or noisy selected-slot fallback tags.

Impact

Readers should see more specific tags led by the main named games, studios, companies, platforms, people, products, or events, while mention-only and rejected subjects stay out of visible card tags. Public tag rows still expose only slug, label, and type.

User action

Nothing.

Added an admin-only read-only scorecard for tag enrichment campaigns plus a local helper that combines saved quality evidence into sanitized issue summaries.

Impact

Operators can see aggregate campaign progress, quality-evidence labels, top remaining failure buckets, and next actions while public tag pages, search, ranking, related stories, rollups, and readiness stay unchanged.

User action

Nothing.

Added an admin-only dry-run/apply lane that can backfill bounded selected archive stories after active dictionary and alias writebacks.

Impact

Operators can preview, enqueue, resume, and process affected story-tag projections while stale guards, retries, audit logging, and display/index/mention-only visibility rules stay enforced.

User action

Nothing.

Added an admin-only dry-run/apply path that can turn safe archive promotion decisions into canonical tag and alias dictionary authority.

Impact

Operators can grow the private D1 taxonomy dictionary with audited, idempotent writebacks while public story tags, readiness, ranking, and tag surfaces stay unchanged.

User action

Nothing.

Added an admin-only deterministic promotion safety report and bounded recompute endpoint for normalized archive AI tag candidates.

Impact

Operators can see which private AI-normalized candidates are promotable, held, or blocked before future canonical writeback work, while public tag payloads and readiness stay unchanged.

User action

Nothing.

Added an admin-only AI archive enrichment campaign runner that can plan and process bounded selected-archive slices for private missing-tag and alias evidence.

Impact

Operators can inspect private archive tag-gap proposals without changing public tag payloads, selected stories, ranking, related-story behavior, or tag surfaces.

User action

Nothing.

Added an admin-only archive scan ledger foundation and coverage report for future tag dictionary enrichment campaigns, with bounded status and rescan-reason tracking.

Impact

Operators can see which selected archive slots are scanned, stale, failed, or still unscanned without exposing private evidence or changing public tag payloads.

User action

Nothing.

Added an operator-only active-learning report for failed tag-quality recertifications, turning failures into bounded cause buckets, fixture recommendations, and safe next-action summaries.

Impact

Operators can create follow-up issues, regression fixtures, and admin-review tasks from sanitized summaries and private local artifacts while public tag payloads stay unchanged.

User action

Nothing.

Platform and storefront tags now distinguish material availability, trial, port, delisting, certification, policy, and exclusivity stories from incidental platform mentions.

Impact

Readers should see platform tags when they explain the selected story, while routine Steam demos, broad launch platform lists, related links, page furniture, and weak storefront mentions stay out of visible story-card tags. Public tag payloads remain unchanged.

User action

Nothing.

Added a reviewed authority cohort for material games, studios, and events that were missing from public story-card tags in the latest certification evidence.

Impact

Readers should see clearer tags on more selected stories that are genuinely about subjects such as Compulsion Games, Double Fine, Ninja Theory, GTA 5, Steam Next Fest, Destiny 2, and related cohort entries, while public tag payloads stay unchanged.

User action

Nothing.

Added bounded recent backfill so newly promoted learned tags can replay recent selected stories and update public tag rows only when current article evidence still passes.

Impact

Readers may see accurate tags appear on recent first-wave stories after safe promotion, while public tag payloads stay unchanged and operators get private dry-run/audit evidence.

User action

Nothing.

Added private admin safety controls for automatic tag learning, including a kill switch, family caps, scoped suppression/quarantine, learned-only tombstone/deprecate actions, exact writeback tracking, and bounded undo for recent automatic writebacks.

Impact

Public tag payloads stay unchanged, while operators can pause, cap, inspect, and undo unsafe automatic tag writebacks without losing private learning evidence.

User action

Nothing.

Added a conservative canary policy that can automatically promote the safest learned game and event tags using private learning, identity-safety evidence, and strict per-story caps.

Impact

Readers may see clearer story-card tags for newly learned safe entities, while public tag payloads stay unchanged and unsafe, ambiguous, duplicate, stale, or review-only candidates remain private.

User action

Nothing.

Added private external identity hints for tag disambiguation so operators can distinguish missing, matching, conflicting, wrong-family, unsafe, duplicate-owned, and unavailable identity evidence.

Impact

Public tag payloads stay unchanged, while future promotion and review work can use bounded private identity evidence without making external providers part of feed freshness.

User action

Nothing.

Added private deterministic learning state, scores, snapshots, and bounded admin review diagnostics for automatic tag candidates and relations.

Impact

Operators can see why candidate tags are being monitored, suppressed, blocked, or considered privately promotion-ready without exposing that evidence in public feeds or tag payloads.

User action

Nothing.

Added private zero-day discovery for new games and named events so selected stories can collect reviewable candidate evidence before any public tag promotion.

Impact

Readers keep seeing only reviewed public story-card tags, while operators get stronger private evidence for newly emerging game and event subjects.

User action

Nothing.

Learned tag aliases now fail closed when evidence is relation-only, wrong-family, short or ambiguous, lifecycle-blocked, stale, or conflicts with existing canonical or curated alias authority.

Impact

Readers should be less likely to see future story-card tags collapse games, studios, platforms, companies, or events into the wrong identity, while private relation evidence stays separate from public tags.

User action

Nothing.

Story cards now keep high-confidence display tags separate from public-safe index rows, mention-only references, and private learning evidence.

Impact

Readers continue to see concise card tags, while operators and future tag features get safer indexed tag rows for discovery without exposing hidden evidence or noisy mentions.

User action

Nothing.

Tightened rolling-feed repeat checks so an already-selected story update cannot be reselected from an older duplicate-chain root or the same canonical URL.

Impact

Readers should be less likely to see the same hot story repeated across adjacent feed slots, while genuinely new material follow-ups can still appear.

User action

Nothing.

Added a local certification report for deciding whether reviewed production public tags meet the roadmap gate before public tag pages, search, related links, rollups, or ranking use them.

Impact

Operators can now produce a deterministic pass/fail JSON or Markdown report with freshness, coverage, privacy, metric, and bounded failure-cause checks before unblocking public tag surfaces.

User action

Nothing.

Expanded deterministic relation learning for clear game, studio, publisher, and ownership stories so safe endpoint tags can appear when the selected article owns the relationship.

Impact

Readers should see clearer tags on more cards where the material subject is expressed through a game-studio or game-publisher relationship, while private relation evidence stays out of public payloads.

User action

Nothing.

Allowed high-signal single-source business strategy, legal, monetization, and company-structure stories through materiality checks when they have strong entity and executive or structural context.

Impact

Readers should be less likely to see the live feed stall during quiet periods when the best available candidate is a material company-strategy story, while broad reports, profiles, and exact repeats remain filtered out.

User action

Nothing.

Added filtered Variety Gaming and Mobidictum intake for entertainment-IP, licensed-game, regional-market, mobile, funding, studio, publishing, and market-data signals.

Impact

Readers should be more likely to see material entertainment-IP and regional mobile-market stories that industry podcasts discuss, while broad Hollywood, recap, review, guide, trailer, discount, and esports noise stays filtered out.

User action

Nothing.

GameDevPress now ingests the Alinea Insight Newsletter as a separate curated industry-analysis source alongside the Alinea Analytics blog.

Impact

Fresh Alinea market-analysis posts can now be fetched, scored, and considered for the selected feed instead of being missed because the newsletter RSS was not configured.

User action

Nothing.

Added a small reviewed authority cohort for material game and studio subjects that were missing from otherwise safe story-card tag extraction.

Impact

Readers should see clearer tags on more cards that are genuinely about subjects like ArenaNet, Gamefam, Alkimia Interactive, Sega, GTA 6, Senua, Clockwork Revolution, Gears of War: E-Day, and Where Winds Meet, while existing noisy-shape abstention guardrails stay in place.

User action

Nothing.

Added filtered Mobile Dev Memo landing-page intake for app-economy, App Store, advertising, user-acquisition, privacy, and monetization signals.

Impact

Readers should be more likely to see high-signal mobile market and app-platform analysis that industry podcasts often discuss, while podcast, newsletter, and evergreen resource links stay filtered out.

User action

Nothing.

Story-card tag projection now abstains from noisy licensed gear, guide, resource, support, roundup, list, quote, column, interview, related-link, weak body-only mention, and generic structural-label shapes while preserving article-owned subjects.

Impact

Readers should see fewer misleading public tags on ambiguous or page-furniture-heavy stories, while legitimate headline, dek, and lead-owned subjects remain eligible.

User action

Nothing.

Added a safe landing-page source adapter for public report and data-insight pages, starting with Niko Partners Data & Insights.

Impact

Readers should be more likely to see high-signal market-model, regional-market, livestreaming, and analyst insight stories that TWIG-style coverage often picks up, without scraping paywalled/private content or broad generic pages.

User action

Nothing.

Selected-feed repeat protection now keeps exact repeats suppressed while allowing true material updates, such as lawsuits, delays, funding, layoffs, closures, regulation, or official responses, to publish as fresh follow-up cards.

Impact

Readers should be less likely to see the live feed go stale during active news cycles where every new candidate was previously treated as duplicate history, while stale reposts and same-URL repeats still stay out of the feed.

User action

Nothing.

Added a local operator dry-run report that replays bounded selected-story tag extraction, compares current and proposed public tag shapes, and separates safe-looking changes from review, stale, blocked-noise, fetch-failed, and no-change rows.

Impact

Operators can inspect exact proposed tag deltas and stable report hashes before any retroactive public tag movement is approved. This first slice does not write production tags or change reader-visible story cards.

User action

Nothing.

Added TWIG-informed source-role scoring and monitoring attribution so platform, policy, market-data, investment, regional-market, entertainment-IP, and analyst-source signals can be evaluated more precisely.

Impact

Readers should be more likely to see material stories from specialist source roles when the article also carries concrete event or market evidence, while operators can separate missing-source gaps from feed-selection and Daily Review promotion gaps.

User action

Nothing.

Added filtered Xbox Wire, PlayStation Blog, Apple Developer News, Android Developers Blog, and Roblox DevForum Announcements intake for TWIG-informed platform, app-store, and creator-economy signals.

Impact

Readers should be more likely to see material platform strategy, app-store policy, Google Play, App Review, Game Pass, State of Play, and Roblox creator-platform changes while routine catalog, podcast, recap, and broad developer-update posts stay filtered out.

User action

Nothing.

Disabled the broad consumer Pocket Gamer feed from active selected-story scanning while keeping PocketGamer.biz available for mobile-business coverage.

Impact

Readers should be less likely to see routine mobile consumer utility stories crowd out material game-dev industry news, while historical archive entries and specialist mobile-business reporting remain available.

User action

Nothing.

Added GamesBeat, Game World Observer, The Game Business, InvestGame, and filtered Circana coverage as TWIG-informed business, funding, deal, and market-data sources.

Impact

Readers should be more likely to see high-signal games-business stories about funding, M&A, restructuring, sales data, units, consumer spending, and market reports.

User action

Nothing.

Low-score single-source business-style cards now need concrete material impact before they can be selected for the public feed.

Impact

Readers should be less likely to see routine reaction, broad market chatter, or consumer-interest items promoted as industry news, while clear business, platform, labor, legal, funding, and developer-impact stories remain eligible.

User action

Nothing.

Added MobileGamer.biz and Naavik as high-signal mobile, liveops, market-data, and games-business sources for selected-story consideration.

Impact

Readers should be more likely to see important mobile UA, liveops, market strategy, and regional-market stories that industry podcasts and analysts are discussing.

User action

Nothing.

Raised Bloomberg Gaming from source-ingestion coverage to explicit feed-priority treatment so eligible Bloomberg game-industry stories can compete as high-score public feed items.

Impact

Readers should be more likely to see eligible Bloomberg gaming coverage in the live feed, including recently discovered Bloomberg items that entered after the source filter was tuned.

User action

Nothing.

Added narrow authority for selected headline-owned game subjects so current and future story cards can show accurate tags for Silent Hill: Townfall and Final Fantasy 11 while preserving tagless abstentions for ambiguous gear, roundup, ranking, and column stories.

Impact

Readers should see clearer tags on more recent story cards without broadening tag pages, ranking behavior, AI writeback, or private taxonomy evidence exposure.

User action

Nothing.

Expanded the filtered Bloomberg gaming source to include Bloomberg's Final Fantasy coverage after the live Technology RSS feed exposed that relevant game-industry stories were being filtered out.

Impact

Readers should start seeing eligible Bloomberg game coverage when Bloomberg publishes it, while unrelated Bloomberg technology stories remain filtered out.

User action

Nothing.

Public readiness and release smoke checks now verify bounded no-candidate droughts against the current target slot instead of selected-story wall-clock age alone.

Impact

Operators can distinguish a short healthy quiet-feed window from stale target slots, over-limit droughts, source starvation, and new-article no-selection failures without exposing private diagnostics.

User action

Nothing.

Added a bounded private admin control that can process selected-story enrichment for exact current story-card targets without draining unrelated enrichment backlog first.

Impact

Operators can safely unblock approved tag verification holds for a small set of current cards while preserving normal scheduled enrichment and existing stale-write guards.

User action

Nothing.

Public readiness and release smoke checks now distinguish a bounded quiet no-candidate feed window from a real selected-story freshness regression.

Impact

Operators can keep release checks green during short, healthy no-new-article droughts while stale, source-starved, new-article, and over-bound no-candidate states still fail.

User action

Nothing.

Selected-story enrichment now recognizes a narrow set of business and operational story-card subjects when the article title clearly identifies a material company, studio, product, tool, or company-game relationship.

Impact

Readers should see clearer tags on some current business and operations cards without adding tags to rankings, roundups, comparison pieces, consumer hardware/legal-tech stories, or outlet staff columns.

User action

Nothing.

Related-coverage clustering now requires more than broad franchise-family overlap before grouping stories as one hot feed item.

Impact

Readers should be less likely to see unrelated articles promoted together just because they share a franchise family name, while true same-story multi-outlet coverage can still receive related-coverage heat.

User action

Nothing.

Tightened hot-video relevance so public hot-video surfaces now publish only videos materially about games or movies, while generic event-wrap, lifestyle, pricing, merch, booth, and other off-topic high-view uploads are suppressed.

Impact

Readers should see fewer but more trustworthy hot videos in the homepage feed and Daily Review, and operators can still inspect suppressed high-view uploads in admin diagnostics without exposing private moderation details publicly.

User action

Nothing.

Added a high-priority Bloomberg gaming source backed by Bloomberg's public Technology RSS feed with a source-level game-industry filter.

Impact

Readers should have a better chance of seeing material Bloomberg coverage about game companies, platforms, and market moves without unrelated Bloomberg technology stories entering the feed.

User action

Nothing.

Expanded low-weight source coverage and selection signals for platform support sunsets, game-discovery infrastructure, live-service end-state stories, and game unit-economics analysis highlighted by the TWIG benchmark.

Impact

Readers should be more likely to see major platform-strategy, studio-impact, market-analysis, and business-model stories even when the first strong report comes from a consumer outlet.

User action

Nothing.

Selected-story enrichment now recognizes more current-article game subjects when headline evidence is confirmed by the lead, summary, or metadata, including demos, trailers, mode updates, named platform-release updates, and newly revealed games.

Impact

Readers should see more story cards name the game an article is clearly about while comparison games, policy examples, creator background, and alias-conflict cases stay out of public tags.

User action

Nothing.

Selected-story enrichment now recognizes more Xbox and PlayStation platform-strategy subjects from current article headlines and summaries while keeping routine catalog, roundup, sidebar, and availability mentions out of public tags.

Impact

Readers should see clearer platform tags on exclusivity and console-availability stories without adding noisy storefront or catalog context.

User action

Nothing.

Selected-story enrichment now projects more canonical company subjects from strong current-article lead context while keeping uncertain company candidates and private taxonomy evidence out of public feeds.

Impact

Readers should see clearer company tags on strategy, publishing, and platform-holder stories without adding source-attribution, background investor, body-only, or company-shaped-title noise.

User action

Nothing.

Tightened live-feed selection again so single-source person, AI, and ad-backlash reaction cards are less likely to win a public slot without concrete studio, developer, platform, legal, or policy impact.

Impact

Readers should see fewer personality-driven AI backlash and brand-ad reaction cards crowding the live feed, while materially important AI, platform, studio, and developer-impact stories remain eligible.

User action

Nothing.

Hot videos now appear inline in the regular GameDevPress homepage feed instead of living in a separate homepage module.

Impact

Readers can catch breakout videos in the same newest-first scan path as written stories, while Daily Review keeps its separate date-scoped video lane.

User action

Nothing.

Improved autonomous story-tag learning for material game subjects while keeping private taxonomy evidence out of public feeds.

Impact

Readers should see more story cards name the game an article is clearly about, without exposing private candidate, review, or relation evidence.

User action

Nothing.

Tightened live-feed selection so single-source showcase reaction, editorial reaction, and thin trailer or mode-reveal cards are less likely to win a public slot without concrete business, developer, studio, or platform impact.

Impact

Readers should see fewer opinionated showcase takes and thin trailer blurbs crowding the live feed, while material platform, studio, developer, and multi-outlet announcement stories remain eligible.

User action

Nothing.

Tightened related-story clustering again so weak shared words and broad showcase or release phrasing are less likely to bridge unrelated game, IP, or company coverage.

Impact

Readers should be less likely to see selected cards mix unrelated stories that only overlap on broad trailer, gameplay, release-window, or common headline wording.

User action

Nothing.

Selected-story enrichment can now safely grow more material event and person tags from current-article evidence, while keeping uncertain products and weak candidates private until stronger evidence exists.

Impact

Readers should see more story cards that name the important subject behind an article without exposing private review evidence or adding noisy broad topic tags.

User action

Nothing.

Daily Review no longer shows or generates the low-quality overview block; story cards keep the short significance line and public-safe rationale cues.

Impact

Readers get a cleaner Daily Review focused on ranked story groups without exposing private scoring weights, source diagnostics, AI prompts, or provider details.

User action

Nothing.

Daily Review now focuses on the top 10 written stories of the day, uses feed-style story cards with visible tags and grouped coverage counts, and summarizes the full top-ten briefing instead of only the lead article.

Impact

Readers can scan the day faster and see why major multi-outlet stories made the list, including how many articles and sources covered each theme.

User action

Nothing.

Added private audited admin actions for single taxonomy candidate decisions, alias decisions, and relation decisions.

Impact

Operators can reject, reopen, block, accept existing candidates, add safe aliases, and curate private relation evidence without exposing private review data or changing public tag APIs.

User action

Nothing.

Improved selected-story tag understanding so GameDevPress can promote more real game and studio subjects from current article evidence while keeping uncertain learned candidates and relation details private.

Impact

Readers should see more article-intelligible story-card tags for funding, acquisition, and launch coverage without exposing private review-queue evidence or incidental storefront/platform noise.

User action

Nothing.

Extended feed repeat suppression to cover a full day of recent selected-story identity, including cases where a resurfaced story arrives with a changed key or missing duplicate attachment.

Impact

Readers should be less likely to see the same story reappear later the same day as a fresh card, while distinct same-franchise developments can still surface.

User action

Nothing.

The homepage can now quietly check for newer selected stories while it stays open and show an inline Load new stories prompt when fresh coverage is ready.

Impact

Readers can notice important new GameDevPress stories without manually refreshing the page or losing their place mid-scroll.

User action

Nothing.

Tightened feed clustering so selected cards are less likely to mix unrelated games or companies that only share broad showcase, trailer, gameplay, expansion, genre, or event terms.

Impact

Readers should see cleaner related coverage on feed cards while true same-story multi-outlet coverage can still be grouped.

User action

Nothing.

Added feed-quality suppression for where-to-watch, liveblog, live report, schedule, and roundup-only event wrapper pages.

Impact

Readers should see fewer selected slots spent on event viewing logistics or broad announcement containers, while material game-dev business, platform, studio, policy, and developer-impact stories from the same events remain eligible.

User action

Nothing.

Daily Review now groups repeated same-day follow-ups into one top story card per major story, adds a deterministic written briefing and theme chips when AI recap output is missing, and moves empty optional video state out of the way.

Impact

Readers should get a cleaner 3-5 theme daily briefing that explains the day without repeated slot variants crowding out distinct stories.

User action

Nothing.

Added a Daily Review page with a date switcher, stored AI recap, date-scoped trending videos, and up to 15 score-ranked written stories from the selected timeline.

Impact

Readers can catch the previous completed UTC day's strongest GameDevPress stories and videos in one compact view without scanning the rolling feed or general Archive.

User action

Nothing.

Added a manifest-driven regression suite that inventories reader and operator-facing features, runs local and live checks, and records report evidence for failures.

Impact

Operators can catch regressions across the feed, archive, patch notes, hot videos, public readiness, admin diagnostics, and Worker jobs on a daily schedule instead of relying only on ad hoc smoke checks.

User action

Nothing.

Added feed-quality suppression for consumer code, promo-code, redeem-code, gift-code, and reward-list utility pages.

Impact

Readers should see fewer selected slots spent on code-list pages, while legitimate game-dev business, platform, SDK, source-code, repository, and code-of-conduct stories remain eligible.

User action

Nothing.

Added private D1-backed selection attempt checkpoints so finalizer work is durably claimed, stage-progressed, and failed closed or completed without relying on one in-memory pass.

Impact

Operators can distinguish active, stale, failed-closed, no-candidate, and selected attempt states while public readiness and feed APIs remain sanitized.

User action

Nothing.

Tightened deterministic private candidate learning so edition labels, legal phrases, funding-company subjects, and company-shaped studio names are no longer queued as actionable game candidates.

Impact

Operators should see a higher-signal private taxonomy review queue without losing material candidate recall or exposing private evidence publicly.

User action

Nothing.

Added a private admin outcome ledger for selected, no-candidate, source-fail-closed, and stale finalizer decisions, with bounded counts for source jobs, ranking, and publication intent.

Impact

Operators can diagnose why a refresh selected a story or failed closed without exposing ledger internals on public readiness, status, latest feed, or Pages surfaces.

User action

Nothing.

Added private admin diagnostics for taxonomy candidate and relation review queues, including bounded queue samples, status/type counts, alias-conflict visibility, and private precision/noise reporting in the local tag-quality gate.

Impact

Operators can judge whether learned tag candidates and relation evidence are useful enough for future curation actions without exposing private evidence publicly.

User action

Nothing.

Added a recent duplicate-chain cooldown so a hotter article from the same already-selected story chain is held back instead of publishing as another public card.

Impact

Readers should see fewer repeated versions of stories already covered in the recent live feed, with the next distinct high-signal story getting the slot when available.

User action

Nothing.

Added compact admin readiness summaries for infrastructure, ingestion/source health, selected-story freshness, and optional background lanes.

Impact

Operators can distinguish stale Worker status, source-starved slots, no-publish outcomes, selected-story freshness, and AI/enrichment/publish/hot-video warnings without exposing private diagnostics publicly.

User action

Nothing.

Bounded finalizer ranking work and moved nonessential cluster persistence behind selected-slot publication and terminal slot recording.

Impact

If a refresh has unusually many story clusters or cluster persistence slows down, the feed can fail closed or publish the selected slot without leaving the current refresh stuck indefinitely.

User action

Nothing.

Archive now shows older selected stories as a flat paginated list of 50 story cards per page instead of a date or range browser.

Impact

Readers can scan past GameDevPress coverage in the same newest-first story order that items leave the latest feed, without date or topic grouping.

User action

Nothing.

Tightened tag aboutness for first-party remake stories so the public tags keep the franchise, platform-holder, and remake topic while incidental competitor availability, engine implementation notes, and generic launch timing stay out of the tag line.

Impact

Stories like a Halo Campaign Evolved remake should show tags such as Halo, Xbox, and Remakes instead of incidental PlayStation, Unreal Engine, or Game Launches tags.

User action

Nothing.

Added a dedicated hot-video feed that can surface recent YouTube uploads from configured channels when they cross an operator-defined public view threshold, starting with IGN.

Impact

Readers can see breakout videos in a separate media module without replacing or polluting the selected written-news timeline. If YouTube credentials, quota, or eligible videos are unavailable, the module stays hidden and the news feed continues normally.

User action

Nothing.

Added Game File and SuperJoost to the curated industry-analysis intake after fresh public-feed checks confirmed they fit the existing RSS size and parsing limits.

Impact

Readers should have a better chance of seeing timely game-dev business, publishing, market, and platform analysis from these specialist sources while normal relevance, duplicate, freshness, and source-health safeguards still apply.

User action

Nothing.

Improved tag selection for remake and remaster stories so the public tags emphasize the game, responsible studios, and remake topic while routine release-platform availability stays out of the tag line.

Impact

Stories like a Hitman remaster announcement should show tags such as Hitman, Saber Interactive, IO Interactive, and Remakes instead of incidental PlayStation, Xbox, or Nintendo availability mentions.

User action

Nothing.

Added curated industry-analysis source treatment for fresh, relevant posts from verified specialist feeds such as Alinea Analytics and GameDiscoverCo.

Impact

Readers should be more likely to see timely market, discovery, publishing, and platform analysis while normal freshness, duplicate, relevance, and source-health safeguards still apply.

User action

Nothing.

Changed the live-backed homepage to show a neutral loading state instead of older static story cards before the live feed finishes loading.

Impact

Readers on slower mobile refreshes should no longer see story cards briefly appear and disappear because they are not in the current live feed.

User action

Nothing.

Added private archive-health diagnostics that summarize selected-story coverage, metadata readiness, and archive authority posture.

Impact

Operators can verify the archive is durable and ready for future bounded consumers without exposing private diagnostics publicly.

User action

Nothing.

Added a public Archive page plus live archive discovery and bounded date-range retrieval for selected stories.

Impact

Readers can browse available archive dates and load selected stories across a short period after they leave the latest feed.

User action

Nothing.

Confirmed the archive reads from the durable selected-story timeline instead of only the latest feed window.

Impact

Readers can still retrieve selected stories by archive date after those stories leave the latest 24-card feed.

User action

Nothing.

Tightened live feed clustering so true related coverage is more likely to support one coherent story while loosely similar roundup, guide, catalog, accessory, or same-platform articles are less likely to merge.

Impact

Readers should see cleaner feed cards with fewer confusing mixed-story clusters and fewer repeated variants when related coverage can safely be grouped or suppressed.

User action

Nothing.

Fixed a homepage refresh flash where unpublished candidate stories could appear before the live feed loaded.

Impact

Readers should only see selected published stories or a neutral loading state while the live feed upgrades.

User action

Nothing.

Added near-term duplicate and repeat suppression for selected live feed cards, including exact repeats, URL repeats, same-source reposts, and near-duplicate cross-source coverage.

Impact

Readers should see fewer repeated versions of the same development across nearby feed cards, while material follow-up developments can still appear as updates.

User action

Nothing.

Improved public story tags for showcase, major update, business, and interview coverage so tags better identify the leading story's actual events, people, games, companies, platforms, and update topics.

Impact

Readers should see cleaner, more complete story-card tags while weak background mentions, platform cautions, role phrases, and page furniture stay out of public tags.

User action

Nothing.

Added feed-quality guardrails so selected stories prioritize higher-signal game-dev industry news over low-value guides, routine catalog churn, thin trailers, and consumer noise.

Impact

Readers should see fewer weak routine cards winning the live feed when material business, developer, platform, studio, labor, legal, funding, or policy stories are available.

User action

Nothing.

Separated backend ingestion and source-job persistence into a dedicated internal Worker store.

Impact

Feed freshness, Queue processing, and source-job recovery are easier to maintain safely without changing the public feed experience.

User action

Nothing.

Moved RSS source refreshes into durable Cloudflare Queue jobs with D1 source-job, attempt, quarantine, and finalizer tracking.

Impact

A slow or blocked source can no longer stall the whole feed refresh path. Each scheduled slot now records terminal source-job state, cleans up stale runs, and fails closed instead of publishing stale articles as fresh updates.

User action

Nothing.

Worker heartbeat AI job processing now honors the documented `DIGEST_AI_JOBS_PER_HEARTBEAT` setting instead of always using a hardcoded single-job limit.

Impact

Operators can tune AI backlog drain rate within the existing safe 1-3 job bound, while missing or invalid values still fall back to one job per heartbeat.

User action

Nothing.

Added a behind-the-scenes tag evaluation gate that scores tag accuracy, missing important tags, noisy tags, alias behavior, promotion behavior, and privacy boundaries against a curated test corpus.

Impact

Future tag improvements now have a measurable quality bar before tags are expanded into more public discovery, ranking, related-story, or recap surfaces.

User action

Nothing.

Moved selected-story AI recap and classification work into durable Worker background jobs, with explicit production enablement and cost-first defaults.

Impact

The public feed can keep publishing even when AI services are slow or unavailable, while operator AI usage is more bounded and predictable.

User action

Nothing.

Added a public Patch Notes page so readers and operators can see recent GameDevPress product, feed, and reliability changes in one chronological list.

Impact

Shipped changes now have a user-safe place to be summarized after they go live.

User action

Nothing.