Changelog

July 1 – August 8, 2026

125 user-facing changes shipped.

New features

• Google News sitemap (48h daily-reflection leaf) (#279)
• owner-gated AdSense footer canary (dark without env) — T3 money-path (#280)
• add image-regen operator lane for artifact_repairable image-gen failures (#278)
• add --drain-backlog lane to run-story-batch for artifact-repairable backlog (#277)
• add Terms of Service page at /terms (#274)
• add consent-gated Google Reader Revenue Manager (Subscribe with Google) (#275)
• stream OpenAI-compatible completions for thinking-model routes (#273)
• add Qwen vision provider to the image text guard (#271)
• enforce explicit News/Reflection/Prayer contract at polish and publish boundaries (#259)
• rescue verse-index backlinks, consent-gated telemetry, UTM allowlist, privacy honesty (#260)
• rescue legacy profiles/comments owner-only read policies as migration 0046 (#257)
• INFO-209 NewsArticle structured data + INFO-208 social sharing UX (#244)
• INFO-207 Resend newsletter integration + INFO-211 support footer link (#245)
• §7 named-victim backstop + React cache() for story lookups (#242)
• enforce News-section floor + purity in polish verdict (#240)
• wire anti-template scaffold lint into polish verdict (#239)
• ISR caching + anti-template lint + prompt quality improvements (#238)
• SEO/performance + content quality improvements (#233)
• stage disabled conditional AdSense boundary (INFO-141) (#232)
• add privacy-minimized individualized outreach register (INFO-188) (#231)
• INFO-148 Phase 2 — Thought dossier contract with stable evidence unit IDs (#224)
• verify publisher attribution (#229)
• INFO-148 content-kind aware theme extractor for Thought lane (#223)
• INFO-146 scored Thought candidate shortlist in dry-run output (#222)
• INFO-150 share card creative system — 4 formats × 2 aspect ratios (#221)
• INFO-149 persist approved text artifacts for image-only retry + INFO-137 RLS verification (#220)
• INFO-136 weekly editorial QA scorecard scheduling (#219)
• INFO-147 growth analytics + INFO-140 chrome measurement contract (#217)
• salvage mobile reader accessibility improvements (PR #165) (#213)
• owner-gated publisher for hand-authored Inspirational Thoughts (#202)
• homepage featured Inspirational Thought hero (#201)
• draft-model split (GLM-5.2 + kimi-k2.6) + internal writer personas (#200)
• env-driven reasoning_effort + output-token floor for OpenAI-compat routes (#197)
• retire Azure entirely; NeuralWatt/GLM 5.2 via OpenAI-compat route (stacked on #194) (#195)
• Thought throughput + story/image quality tranche (reopen of #194) (#196)
• improve Thought quality recovery (#191)
• launch image-safe reader feed (#170)
• add inspirational thoughts lane (INFO-122) (#143)
• add source health report (#142)
• gate daily-cap increases with a ramp report (#141)
• diversify batch selection and ramp cadence (#138)
• add durable IndexNow outbox (#136)
• make checkpoints durable (#137)
• finish faith-era launch ticket sweep (#128)
• migrate Bible translation from ESV to NLT (#127)
• section pages real-data rebuild — 6 working sections replace dead person-category query (#115)
• conflict-casualty intake guard + stop re-selecting content-rejected events (#113)
• person-less story path + faith-pivot RSS diet trim (#111)
• adopt the Axiom SoT-wiki structure for Infolitico (#109)
• commentary quotes — verified theologian quotes linked to Bible verses (#107)
• friendly-pointed satire register + governance lockstep (#102)

Updates

• add structured failure logging to the Reddit feed scraper (#285)
• Add owner-gated Qwen 3.8 bulk-generation harness scripts (#247)
• Record INFO-197 deployment receipt (#253)
• Prevent hardcoded Scripture drift (INFO-197) (#252)
• Add guarded final-tier generated image library (#250)
• Add fail-closed Qwen streaming draft provider (#249)
• INFO-201: Phase-5 thought-dossier hardening + LLM citation extraction (#243)
• INFO-186: prepare reviewed On Mission paid-seed packet (#228)
• INFO-181: enforce canonical outreach attribution (#227)
• Ship reviewed INFO-152 organic share asset baseline (#226)
• INFO-139: record conditional AdSense-only decision (#225)
• Harden generated imagery against synthetic crowd tropes (#215)
• INFO-136: add read-only editorial repetition telemetry (#214)
• Theme-keyed curated photo pool (INFO-177) (#207)
• Commons relaxed relevance stage-2 + yield metrics (INFO-176) (#206)
• Image aesthetic diversity: scene memory, palette signature, cluster gate (INFO-175) (#205)
• Quality batch: stories, images, UX (INFO-161..173) (#204)
• INFO-129: Gate monetization on Bible-rights readiness (#189)
• Repeat the final length contract in the polish payload (#187)
• Make Thought polish compress overlong intermediate drafts (#186)
• Let the final guard enforce Thought publication length (#185)
• Bound the Thought compression retry output (#184)
• Compress the actual overlong Thought draft on retry (#183)
• Harden the controlled Thought operator path (#182)
• Keep the new Culture source tier operationally green (#181)
• Expand Faith & Church and populate Culture (#180)
• Record live Reddit feed release (#179)
• Keep Reddit feed posts on Infolitico (#178)
• Record first Thought and live image remediation (#176)
• Retry overlong Inspirational Thought drafts once (#175)
• Fix one-shot recovery for grounded Thought outlines (#174)
• Keep social images in sync during remediation (#169)
• Harden historical image remediation retries (#168)
• Prevent irrelevant political imagery on community stories (#167)
• Fix social card font rendering (#164)
• Harden event image relevance (#163)
• Wire portrait semantics into the live homepage (#161)
• Restore relevant person-first story imagery (#160)
• filter unusable source summaries from batches (#159)
• fix Reddit source grounding (#157)
• fix release safety and publish local work (#156)

Fixes

• fetch each Feed community individually; parse both Reddit title markup variants (#286)
• match Reddit's redesigned embed title markup on The Feed (#284)
• prefilter pre-contract polished bodies in image-regen lane (#283)
• sitemap health check expects sitemap-news.xml leaf (#282)
• allow AdSense sodar hosts in CSP (violation-evidence) (#281)
• splice curated verse into the blockquote so GLM-5.2 can't corrupt it (#276)
• punch list — single prayer regex, line-level legacy split, multi-paragraph prayers, consent test (#269)
• hero renders story primary image eagerly; notice banner becomes docked bottom bar (#254)
• de-tautologize Commons relevance gate + geography guard (#241)
• enforce Scripture quote provenance (INFO-180) (#234)
• enforce isolated runtime smoke (#230)
• INFO-158 section pages lead with visually strong story (#218)
• suppress duplicate decorative quotes when Scripture owns the boundary quote (INFO-101) (#212)
• broaden image-scene abstraction motifs beyond table-and-desk (#211)
• serialize Commons API pacing for apply-time refresh (#210)
• pace theme-photo downloads with 429 backoff (#209)
• theme-photo seed search query sparsity + pacing (#208)
• retry generated images with guard feedback + harden people scenes (INFO-160) (#203)
• 0035 backfill skips legacy non-URL source refs (as applied to prod) (#199)
• restore /story/<slug> permalinks (#192)
• image every homepage hero story (#172)
• image secondary homepage cards (#171)
• make archive discovery mutation-safe (#151)
• consolidate crawl, discovery, and reader polish (#149)
• overscan/paginate batch selection past blacklisted rows (#125)
• stop Comments crashing SSR when Supabase env vars are unset (#124)
• diversify outline.md headline examples to break formula repetition (#117)
• generate story-grounded image scenes via LLM instead of canned theme library (#116)
• run topic-dedupe for person-less faith-contract stories (#114)
• normalize markdown chrome out of story bodies + LLM empty-content telemetry (#112)
• thread polish gate scores + retire negative-frame headline gate (#110)
• correct 3 bible_verses with genuine ESV wording drift (#108)
• polish reject-verdict parsing + faith RSS source expansion (#106)

August 15, 2026

Documentation expansion

• Added a reader guide for sections, stories, people hubs, and the archive.
• Added a The Feed guide covering curated public snapshots, disclosure, discussion sign-in, and the shared edition.
• Added a policies guide covering AI disclosure, corrections and retractions, community standards, privacy, terms, and the About page.
• Clarified in Get started that browsing needs no account while Feed discussions require reader sign-in.
• Expanded the morning newsletter guide with cadence, content, consent, availability, and archive guidance.