Product Updates

Released updates and work in progress.

21 release entriesCurrent v0.13.4Released and in development
NEXT
IN DEVELOPMENT

Models and Companies Directory UX

Find models with combined filters and reach company coverage, product links and field evidence more directly.

New Capabilities

Model search with combined filters, result counts and individual removal Compact model lists with access, weights, license and a release shortcut Company profiles and latest coverage have separate links Detail navigation, consolidated product links and evidence-first model access

What Changed

  1. Models keeps search and filters in the URL. Companies separates browse view and sort, and labels the action that applies filters.
  2. Company product entries group official links, documentation and sources. Model detail puts verified access and license before specifications.
  3. Planned follow-up work includes model comparison, additional source coverage and growth experiments; these are not part of the v0.13.0 release.
v0.13.4
LATEST RELEASE

Company Intelligence Directory

Global company changes, compact discovery, exact-name search, business profiles and field-specific evidence with official access links.

New Capabilities

Recent changes are ordered across industries, with an optional industry view Compact mobile filters, contextual counts and independent filter removal Exact company/product matches rank first with escaped Unicode-aware highlights Company-level focus and representative products replace first-product summaries Named fact-check dates, primary/secondary provenance and review-due labels Official sites, API documentation and model-specific license/access sources

What Changed

  1. The directory keeps all 77 profiles and their URLs. Default recent-change order is global; exact-name search is prioritized within the selected order. Profiles without published Events remain explicit.
  2. Reviewed 23 company fact fields across 16 actors through audited admin writes. Added language, multimodal-understanding and document-recognition discovery, and reviewed company-level focus for key profiles.
  3. Added XtalPi and VeriSilicon product entries, updated 4Paradigm's official platform entry and corrected WALL-A from a robot to an embodied model. Overseas access is never inferred from a product link.
  4. Reviewed profile-only coverage against private signals, triage and configured sources. Keyword mentions and similar names do not become company associations, and no sources or blocked Events are automatically promoted.
  5. Existing editorial hardening preserves deterministic split URLs, conservative model-topic matching and digit-separated model-name checks. Zod is updated to 4.6.7 with the committed dependency security fix.
v0.13.3
RELEASE

Company Discovery & Evidence

Company and institution filters, latest verified changes, official product updates and four new China-linked company profiles.

New Capabilities

Companies, research institutions and regulators have distinct filters and counts Agent, coding, search, office and inference discovery with reviewed API tags Latest public Event dates, name sorting and per-field fact-check dates Compact capability selection on mobile Source-backed company identities and official product links

What Changed

  1. Added LangGenius / Dify, VAST / Tripo, InfiniFlow / RAGFlow and Meshy, bringing the directory to 77 profiles: 72 companies, two research institutions and three regulators.
  2. Updated official product entries for Shanghai AI Laboratory, BAAI, NetEase Youdao and Galbot. Product focus descriptions link back to sourced offerings on company pages.
  3. Company cards show the latest readable Event in the current public projection. Sorting uses occurrence dates within each category; a field check is not presented as a whole-profile review.
  4. Official company facts are reviewed field by field; unsupported historical values keep their previous verification dates. Overseas locations and China operations are stated separately, and overseas profiles no longer default to a China address in structured data.
  5. New profiles do not activate sources, relabel historical Events or bypass editorial gates. Registry maintenance remains in shadow mode.
v0.13.2
RELEASE

Event Integrity & Registry Reconciliation

Original-evidence clustering, audited association repairs, corrected Event dates and verified model and company updates.

New Capabilities

Clustering uses original anchors and separates distinct assets, versions and changes Later sources preserve occurrence and publication dates Evidence repair supports previews, concurrency checks, audit and safe re-triage AI relevance is checked before clustering, drafting and publication Reviewed model-to-Event links require subject and official release evidence

What Changed

  1. Repaired 298 unrelated associations across 82 Events and corrected historical occurrence dates without deleting source material or changing Event URLs.
  2. Withdrew an unrelated NMR instrument story and a funding story whose remaining source is still in shadow verification; their evidence remains private.
  3. Added published profiles for LongCat-2.5, Xiaomi-OCR-0 and Qwen-Image-2.1-Turbo. MiDashengLM-Spatial remains a draft until its official release date is verified.
  4. Reverified 198 existing model fields against official cards, public weight manifests and license text, including commercial and territory conditions. Unverified historical fields retain their previous dates.
  5. Updated 19 facts across eight companies and the Xiaomi and Meituan product entries. MiniMax's founding year now follows its official company introduction.
  6. Registry maintenance remains in shadow mode. Detached reports return to normal review and are never published merely because a repair was approved.
v0.13.1
RELEASE

Production Reliability

Stricter context evidence, accurate publication monitoring, independent Cloudflare health checks, bounded source recovery and private-page analytics protection.

New Capabilities

Context counts must match an adjacent label, number and unit in the cited text, including when applying an existing candidate Publication freshness uses the latest visible publication time, including newly published historical Events Hourly Cloudflare health checks persist safe statuses and expose stale monitoring through the public health API Bounded WordPress listings support nested fields, explicit UTC dates and AI keyword filtering Private HTML prevents CDN transformations while retaining its existing Content Security Policy Newsletter publication delays are visible without changing original dates or publishing a new issue

What Changed

  1. Rejected an unsupported proposal to change DeepSeek context from 1M to 200K; verified public facts remain intact and maintenance stays in shadow mode.
  2. Fixed the monitor API-shape mismatch that reported an empty catalog despite published Events. External summary failures no longer prevent the health probe from running.
  3. Added idempotent hourly production checks and explicit stale-run reporting independent of GitHub runner billing. The external GitHub mirror remains subject to its account limits.
  4. TechNode uses a small public WordPress listing in shadow verification. Source reconfiguration is restricted to the reviewed catalog and requires authenticated approval with audit; Alizila remains quarantined while returning HTTP 403.
  5. Member pages send private, no-store, no-transform to prevent injected analytics. Newsletter pages mark a public issue older than eight days as overdue and preserve its real publication date.
Earlier releases (16)
v0.13.0
RELEASE

Models, Companies & Indexing

Evidence-backed model and company maintenance launches in experimental shadow mode, with reviewed evidence, current official product links, honest Event URL states, and improved newsletter reading and search entry points.

New Capabilities

Official-evidence registry proposals, a 30-day backfill queue, field diffs, approval and conflict-safe rollback; production runs in experimental shadow mode Audited evidence attachment and evidence-dated corrections for existing Events, with published projection refresh and duplicate-clustering prevention Company profiles show verified models alongside recent release coverage and current official product and API links Historical Event URLs preserve honest 404 states and retryable 503 responses for temporary read failures Newsletter pages expose recent issues and archives, and explicitly cited Events link to their newsletter coverage Shared indexing eligibility and complete model-fact reads keep public HTML and sitemaps consistent

What Changed

  1. Manual evidence review: approvers can attach reviewed evidence to existing Events and correct a date only against matching dated evidence, with audit and public projection refresh. Company profiles add official DeepSeek Harness, Kimi Code and MiniMax Code/API links; the Kimi API link points to the current international platform. Model facts are updated separately from future release promises and historical reporting.
  2. Registry maintenance (experimental; initial rollout uses shadow): official-evidence proposals for DeepSeek, Qwen, Kimi, GLM and MiniMax, a 30-day backfill queue, field-level diffs, approval and conflict-safe rollback. Only deterministic public-weight releases qualify for automatic entry after rollout verification. Company pages show verified model profiles alongside recent release coverage.
  3. Retired Event URLs now return an honest 404 with reading links instead of redirecting to the homepage. Temporary data failures return a retryable 503 rather than a permanent removal or redirect.
  4. Newsletter reading paths now include up to three recent published issues with dates, summaries and direct links, plus a separate archive link. Events explicitly cited by an issue link back to it. Feed failures keep the introduction and archive available with a short cache.
  5. Model directory indexing fixes: published release coverage keeps the directory eligible for search even before model profiles are ready. Complete batched model-fact reads keep company and model sitemap eligibility consistent as the registry grows.
  6. Homepage subscription dedupe: the first-screen call to action stays Get the China AI brief, while the latest-issue card uses Subscribe and Read the latest issue. The subscription destination, analytics event and RSS hydration stay unchanged.
  7. Indexing integrity update: published, evidence-backed Events with complete analysis no longer require launch keywords or an extra 80-character analysis minimum. Empty brief lists, identity-only company profiles, models without substantive verified facts and empty model/policy lists stay readable but out of search and sitemaps; sample pages gain canonical URLs. Google indexing and traffic effects remain unverified.
  8. Newsletter search entry update: the in-site page identifies its China AI newsletter focus, explains coverage and evidence, links to models, companies, policy and Event history, and includes publication structured data. Article reading CTAs lead to this page; issue and subscription links still go to Substack.
v0.12.0
RELEASE

China AI Focus

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

In-site newsletter page: /newsletter/ now returns a real page — value pitch, latest public issue hydrated from the Substack feed, Free / Paid ($10 per month, $100 per year) / Founding ($200 per year) plan cards stating that public Events and evidence stay free, FAQ, and one-hop subscribe links with UTM — instead of a 302 to Substack; /newsletter/index.html and the retired /en/ mirrors 301 to the in-site page while /p/, /s/ and /archive/ handoffs are unchanged Header and homepage conversion surface: the anonymous header shows Start free as the primary member action with a quiet Sign in beside it, the Newsletter nav entry points at the in-site page, and the homepage masthead gains a value line, a Get the China AI brief CTA and a latest-issue card hydrated from the same feed with a capped side-load so a slow Substack fetch can never stall the homepage Paid-funnel analytics: public pages emit newsletter_view, subscribe_click and start_free_click through GA4 consent mode via a fixed-vocabulary funnel helper with click delegation; paywall_view, checkout_start and paid_active stay reserved until an in-site surface exists, and no email, IP, user agent or member id is attached Publication stagnation alerting: the daily monitor checks the newest published Event age on the public API (warning at 72 hours, critical at 7 days) so editorial runs that complete without publishing can no longer pass silently, and the alert decision schema now carries the drafting and publication checks into fingerprints and model review Admin unpublish UI: the event.unpublish action (published -> hidden plus public-projection delete, 2026-10-12) finally has a console button — approvers can retire a published Event with a reason and audit trail instead of calling the API by hand Disaster-recovery source restore: activate-shadow-batch now reads the China catalog and understands all six collector adapters, and a new --restore-active mode emits lifecycle=active recovery SQL with dr-restore-script audit rows from a slug list, so a rebuilt D1 no longer strands 74 active sources with no script to restore them Shared LLM budget for the newsletter pipeline: newsletter generation and admin section rewrites join the editorial receipt book — identical retries reuse the paid result and the shared daily budget throws llm_budget_exhausted instead of quietly paying again Public API documentation alignment: openapi.json grows from 8 to 19 documented endpoints (home, event detail, signals, trend subroutes, sources, attention, newsletter/latest, pro-samples, itself), fixes the limit contract (1-100, default 24), drops the nonexistent events actor parameter and renames the models lab filter; the indexable /briefs/ sample page joins the sitemap instead of sitting outside it Admin ops viewer: a read-only Runs and Audit tab exposes editorial job summaries (selection, drafting and publication counts per run) and the audit log with action-prefix filters — previously only direct D1 reads could answer them, which blocked the selection-and-distribution O2 acceptance check Conditional GET for the public JSON API: every 200 JSON response carries a strong SHA-256 ETag and If-None-Match (weak tags included) now answers with an empty 304 that keeps Cache-Control, so agents, feed readers and CI probes stop re-downloading unchanged payloads Browser smoke in CI: a Playwright Chromium suite boots the Worker under wrangler dev on an empty local D1 and pins the conversion surface (masthead CTA, latest-issue card, header Start free and Sign in, in-site newsletter nav), the in-site /newsletter/ page structure with plan cards and UTM subscribe links, the legacy path handoffs, and the public API contract including the openapi ETag 304 — turning the manual browser-smoke acceptance step into a CI gate (e2e.yml) Admin event list and batch actions: a console tab lists Events by status and keyword instead of requiring hand-typed IDs, and checked Events run the exact single-event visibility transitions in bulk (hide, restore, redraft, unpublish) with per-event audit rows marked batch, one shared reviewer reason, projection deletes on unpublish, and status-mismatched rows skipped and reported instead of blocking the batch — capped at 200 Events per call Draft number checks extended beyond the fact summary: bare numbers in the summary and both insight fields must now trace to the evidence text, while model identifiers glued from letters and digits (V4, Qwen3.8, GPT-4o) and standalone years count as phrasing and the forward-looking fields stay unchecked Admin entry routing tests: the admin worker entry grows from 3 to 9 cases covering the dev identity header, single-user ADMIN_TOKEN accept and reject, unknown API paths answering 404 instead of falling through to assets, JSON content-type preflight on mutations, cross-origin rejection when ADMIN_ORIGIN is pinned, and the query-token to HttpOnly session-cookie shell flow Manual copy editing for Events (editorial-throughput T12): approvers can edit the seven public English fields of a drafted Event directly — with the drafting gates enforced on the admin side (field bounds, English-only, placeholder and shared banned-phrase checks, factSummary preamble, and a Reportedly: hedge that cannot be edited away), an event.copy_edit audit that records only changed fields, an unchanged slug, and a single-row active-version re-projection for published Events; no model call, no drafting attempts, no LLM budget Clustered write batching: the three per-signal attach writes (event_signals, signal_triage, event timestamps) now run as one D1 batch, deferred triage rows flush 25 at a time as idempotent upserts, and the post-attach evidence read is shared between rescoring and the evidence policy — cutting per-signal D1 roundtrips from roughly 5-7 to 3-4 per hourly clustering run with the findMatchingEvent read path unchanged Editorial backlog ceilings: the selection and drafting candidate SELECTs, previously unbounded so query cost grew with review backlog, are capped at four times the per-round paid-call maximum and 200 rows; processed rows leave the candidate set so overflow waits for the next hourly run with per-round ordering unchanged Dead code and dependency hygiene: the unused applySourceSuccess/Failure/canRunScheduled shadow health model with stale thresholds is removed along with three orphan fixtures, the private worker isoWeek duplicate moves onto the shared helper, github-releases gains its first contract test from the real GitHub Atom fixture, and CI now runs npm audit (high/critical, runtime deps) plus a weekly Dependabot schedule for npm and GitHub Actions Funding-beat source coverage: Zhidongxi (zhidx.com/rss) joins as the seventh Chinese Tier-2 media source in shadow with the same AI keyword filter, closing the funding and corporate-scoop gap left by 36kr (volcano-engine WAF on every endpoint, never bypassed) and LatePost (robots.txt disallows all crawling plus a paywall); its first feed window already carries the 3.3-billion-yuan Manus round led by Tencent Homepage daily freshness by design: the More China AI news list now carries every published Event, including ones the selection scoring rejected, while the top-three brief stays selected-only, so every publishing day is visible on the homepage; promotional items that were auto-hidden never return to any list Selection thresholds gain an AIHOT-aligned Tier 1.5 at 65 for expert-creator evidence (industry newsletters such as ChinAI and Tech Buzz China) between official Tier 1 sources (54) and independent media (76), so expert-sourced Events no longer face the media bar Fewer stalled drafts and a material brief: media-only drafts that omit attribution get a deterministic ', outlet reports' fallback instead of burning all four drafting retries, and the homepage material-change gate now recognizes completed funding rounds, IPO preparation, model and provider switches, so real capital and switching stories can lead the brief; Hugging Face release drafts may state the labeled publication date instead of burning retries, and CJK-headline Events get a proper English-title URL slug instead of the event- placeholder Old Event URL handling: every /events/{slug} that no longer exists now returns a permanent 301 to the homepage, and previously unpublished Events no longer get a temporary 302; currently published Events are unaffected and still return 200 Company and product letter marks: every company card, company page header and product card carries a deterministic monogram in the site palette (first letter of the name, ink on soft paper with a hairline border) instead of third-party logos, so all 73 profiles and their products have a visual identifier with no trademark assets to license, hotlink or store Company profile layout redesign: the detail page moves to a two-column layout on desktop with Products, Models, media and Latest Events in the main column and the Verified facts record, official channels and profile sources in a side column; each product renders as its own card (name, Chinese name, source link on the head line, one-line introduction below) grouped under kind headings, and the Known-for line trims to three names with a +N count instead of one long run Company products, grouped and introduced: every catalogued product now declares a kind (model family, model API, assistant app, image, video, 3D, speech, AI search, office AI, robot, chip) with a one-line introduction, and company pages render the product section grouped by kind instead of a flat link list; model APIs count as products, so DeepSeek, Zhipu BigModel, Kimi, Alibaba Bailian and Volcano Ark appear as API entries beside their chat apps AI company directory expansion: thirteen more companies cross-checked against the 2025 Hurun China AI Top 50, the 2026 Forbes China AI Top 50 and comprehensive-strength rankings — JD.com (JoyAI), Qihoo 360 (Nano AI and 360 Zhinao), Kingsoft Office (WPS AI), Megvii (Face++), CloudWalk, 4Paradigm, MetaSo, XtalPi, Genspark, Horizon Robotics (Journey), Rockchip, VeriSilicon and Hygon (DCU) — with official channels verified per company, growing the registry from 60 to 73 entities; Quark and Yuanbao join Alibaba and Tencent as ambiguity-guarded product aliases Watchlist followed-company rows are readable again: the Open Event and Company page actions now share one stacked action cell instead of both being pinned to the same grid cell at desktop widths and rendering on top of each other; the desktop column rule targets only direct-child row links Design-system enforcement and readability: public pages drop all hover lift translations and glow shadows, flatten surviving rounded cards and pill chips (event sidebar, source directory, company rail, tag chips), tokenize 10px micro-labels back to the type scale, and raise semantic labels plus mobile navigation to the --text-xs floor; the admin console is re-skinned on the ink/vermilion tokens (AIGC.NEWS wordmark, Chinese section labels, zero-radius panels, 12px text floor) with persistent plain-language signed-out state instead of a raw HTTP 401 toast, ARIA tab semantics with arrow-key movement, and a fix for the china console sources tab that never revealed its panel Honest empty and failure states: the homepage renders an explicit empty-brief notice and withholds the share action when the projection has zero Events, /lines/ states plainly that line reviews are unavailable with paths to Event History and Signals instead of a false refresh hint, and Event detail pages load the shared page-polish stylesheet for parity with static documents Loading and security hardening: tokens.css inlines the brand font-face and palette at export time to remove the render-blocking import chain, Worker-rendered HTML and the static export send CSP, nosniff and Referrer-Policy headers, the light paper theme becomes the CSS root default so a missing data-theme attribute can no longer flip a page dark, the theme toggle storage key is renamed ainews-pulse-theme, and the changelog page discloses older releases behind a collapsed section Member surface and copy cleanup: the sign-in page ships a single stylesheet after removing the fully overridden legacy layer, the share page drops its pre-override blue CSS, the public Pro sample page is re-skinned on brand tokens with sentence-case labels, dev checkout pages join the brand palette, company profile source links are named by host instead of "Source 1", the footer catalog stamp reads "Catalog updated" to stay distinct from the homepage brief stamp, and duplicated "Suggest a source" card copy is de-duplicated Page favicons: static, dynamic, sample, member and admin console pages share the brand icon; both public and Account hosts serve the same-origin SVG, with a legacy /favicon.ico redirect and ink/vermilion colors Newsletter entry and compact footer: Policy moves to More while remaining in the footer, Newsletter menus open AIGC Newsletter directly, legacy /newsletter/ routes bypass old page caches and use a fixed-destination 302, subscription links open Substack /subscribe, and static and dynamic footers share a 64rem layout with grouped navigation (AIGC Weekly remains a publication section) Member visual consistency: Account, Watchlist, sign-in, plan/billing and retained member tools share the public brand fonts, wordmark and warm-paper/vermilion palette, with a compact 64rem layout and horizontal member navigation; self-hosted fonts also work on the separate Account domain, with unchanged auth and billing gates Compact reading toolbar: the homepage Watchlist and Newsletter entry uses a body-font heading and equal text actions instead of filled and outlined buttons, with responsive stacking, 44px targets and unchanged member-session behavior China AI positioning: registry of 29 Chinese AI entities (labs, big tech, robotics, apps, chips, regulators) and deterministic China nexus classification (production data not yet reset) Company directory discovery: 49 registered entities including 15 added companies in creative AI, speech, inference services, chips and robotics; sourced product links, Chinese/product search, capability filters and real published-Event counts distinguish profile-only entries from coverage (no new collection sources enabled) Company search cards prioritize the matched product within the three-product limit and display verified Chinese product names, so searches for Jimeng or its Chinese name visibly identify the right company China context on Events: nexus-gated drafting and publishing, evidence-bound availability and claim labels, translation notes, and audited editor overrides China source catalog: new model releases from 15 Chinese AI organizations on Hugging Face, lab blogs, English China-tech media and newsletters, with robots.txt enforced on every fetch and audited manual evidence entry (sources start in shadow) China AI navigation: Today and Newsletter up front, AIGC Newsletter page, and method and copyright pages that explain China nexus, translation, outside-China availability and claim status Chinese AI models, companies, policy and explainers: evidence-backed model registry with outside-China availability and open-weights filters, company profiles, a policy timeline and five sourced explainers, plus /api/v1/models and /api/v1/companies Weekly price watch skips non-USD list prices: model prices without dollar amounts (per-second, per-credit and per-resolution video and image plans) no longer go through dollar-amount matching, are listed openly as skipped with a non_usd_price reason, and their evidence pages stay unfetched, so the recheck summary and the weekly brief price section no longer report meaningless matches China AI Lines: five long-running stories (open weights, price war, robotics, chips, regulation) with working theses that start as insufficient evidence and are revised from verified Events Follow Chinese AI companies: Follow on company pages and a Watchlist panel with each company's latest verified Event Sponsorship policy and labeled sponsor slots on the site and in AIGC Newsletter, with sponsor disclosure on coverage and aggregate click counts only (off by default) Ink and vermilion theme: warm paper, ink text and a vermilion accent in light and dark mode, all text at WCAG AA contrast Site-wide reading polish: compact headings, ruled company and explainer indexes, visible model filters with All models, responsive company navigation, matching member colors and accessible admin controls Clearer copy across reading and supporting pages: selected stories with live totals, source updates distinguished from verified facts, dated archive samples, precise model and company introductions, a simpler Watchlist empty state and consistent method, policy and release-note reading layouts Reader-first content hierarchy: concise page introductions, a collapsed homepage reading index, a single Watchlist value statement and newsletter samples before optional reading rules; evidence, dates and uncertainty remain visible Homepage redesign: China AI masthead, company rail linking DeepSeek, Qwen, Kimi, GLM, MiniMax and more, and Explore cards for models, companies, policy, attention and explainers SEO for China AI: keyword-rich titles, descriptions and structured data naming Chinese AI companies and products on every page, with Event-to-company links China-only public site: static snapshot pages, timeline, homepage, signals and sitemap keep only Events and signals with a China nexus; the source map lists the China catalog English newsrooms of Chinese AI companies: DeepSeek, Alibaba, Baidu ERNIE, Unitree, Manus and SenseTime official English news collected as Tier 1 evidence (sources start in shadow) China Attention Index: weekly attention to Chinese AI companies in English- and Chinese-language sources, with evidence links, reviewed before publishing (off by default) Site-wide fixes from a reader walkthrough: Models, Companies, Policy, Explainers and sitemaps route to the Worker, a working 404 page, model releases while the model table fills, more-from-company and model links on Events, company products and website, China-only source counts and company coverage, Follow that completes after sign-in, and Attention Index links only once a week is published Images and video on Events: official images hosted on AIGC.NEWS with credit and source links, YouTube and Bilibili videos that load only when you press play, real share images and VideoObject data, homepage thumbnails and Latest media on company pages (off by default) Free Event followups: factual revisions, evidence changes and clearly labeled related developments, with explicit review state (rollout disabled) Selection and distribution: two-pass selection scoring with tier thresholds, answer-first writing rules, RSS feeds with an agent guide, and receipts with a daily budget for paid model calls Reader surface: stories on the first screen, registry company names, type filters, next reads on Event pages and edge-cached public pages Chinese official sources: DeepSeek, Unitree and Zhipu announcements in Chinese and CAC AI rules and filings, collected as Tier 1 evidence (sources start in shadow) Editorial throughput: drafting backoff, one targeted correction, model-card and README evidence, a 60-day history cut-off and model-version-aware clustering China source expansion: 74 sources across Hugging Face, official GitHub organizations, lab blogs and Chinese and English media, with AI keyword filtering for general tech outlets and China-only public signals Live source health: the Worker reconciles every source daily (degrade, quarantine, recovery probes), /sources/ shows live lifecycle and health, and /api/v1/sources publishes an allowlisted health report Billing catalog freshness: the member Worker's hourly Cron re-verifies each configured Stripe price against the provider (at most one request per offer every 12 hours) and stores the outcome in D1, so offers no longer silently expire 24 hours after a deploy, and a provider/config price mismatch or missing price immediately withdraws the affected offer regardless of the deploy timestamp; a changed price never inherits the old price's verification, renewal and cancellation webhooks map prices without depending on catalog freshness, and every Stripe call has a 10-second timeout (production Stripe not configured)

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.11.1
RELEASE

Audited Research Continuity

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.11.0
RELEASE

Clearer Evidence and Automated Briefs

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.10.0
RELEASE

Living Evidence Interface

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.9.0
RELEASE

Evidence-Dense Decision Intelligence

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.8.1
RELEASE

A Denser Action Surface

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.8.0
RELEASE

The Industry Evolution Map

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.7.0
RELEASE

The Autonomous Intelligence Loop

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.6.0
RELEASE

The Intelligence Atlas

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.5.1
RELEASE

Snapshot Parity

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.5.0
RELEASE

The Evidence Engine

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.4.0
RELEASE

The Industry Arc

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.3.0
RELEASE

Primary Source Protocol

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Evidence-led event coverage Bilingual publishing controls Source and signal monitoring

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.2.0
RELEASE

The Source & Scout Foundation

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

SourceRun Opportunity Scout

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.
v0.1.0
RELEASE

First Signal

An editorial intelligence update focused on clearer evidence, stronger localization, and a more reliable public reading experience.

New Capabilities

Control Room

What Changed

  1. Improved the public reading experience and strengthened the underlying editorial workflow.