Skip to content

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary

fix(digg): search AI and Tech sections and surface a stale Digg copy (#1320)

Matt Van Horna96086d227b6
diff --git a/CONFIGURATION.md b/CONFIGURATION.md
index 88d44b8e..ea82ec94 100644
--- a/CONFIGURATION.md
+++ b/CONFIGURATION.md
@@ -173,7 +173,7 @@ The source table covers the engine's research sources and optional comment or tr
| YouTube comments | `yt-dlp` CLI installed — **free and keyless, no API key and no opt-in needed**. Falls back to `SCRAPECREATORS_API_KEY` + `INCLUDE_SOURCES` containing `youtube_comments` only when yt-dlp is absent. Suppress with `EXCLUDE_SOURCES=youtube_comments`. | top comments (by likes) on the top ~3 videos by engagement | yes — free via yt-dlp (no credits spent) |
| TikTok comments | `SCRAPECREATORS_API_KEY` + `INCLUDE_SOURCES` contains `tiktok_comments` (**on by default** — Step 5 Recommended tier) | top comments (by `digg_count`) on the top ~3 TikTok posts | ~3 calls/run; 10K free calls |
| Instagram comments | `SCRAPECREATORS_API_KEY` + `INCLUDE_SOURCES` contains `instagram_comments` (**on by default** — Step 5 Recommended tier) | top comments (by `comment_like_count`) on the top ~3 Instagram posts, via `/v2/instagram/post/comments` | ~3 calls/run; 10K free calls |
-| Digg | `digg-pp-cli` on PATH (auto-installed during first-run setup via `npx -y @mvanhorn/[email protected] install digg --cli-only`; binary defaults to `$HOME/.local/bin` — Hermes/OpenClaw agent subprocesses must inherit that dir on PATH for Digg to activate; prior pp-digg installs use the same path) | always on if `digg-pp-cli` on PATH | yes (free, keyless, read-only) |
+| Digg | `digg-pp-cli` on PATH (auto-installed during first-run setup via `npx -y @mvanhorn/[email protected] install digg --cli-only`; binary defaults to `$HOME/.local/bin` — Hermes/OpenClaw agent subprocesses must inherit that dir on PATH for Digg to activate; prior pp-digg installs use the same path) | always on if `digg-pp-cli` on PATH; searches Digg's `ai` and `tech` sections. Run `digg-pp-cli sync --topics ai,tech` (about 2 minutes) for full 30-day coverage; without a sync in the last 6 hours, Digg searches only its newest stories and the brief marks Digg partial | yes (free, keyless, read-only) |
| arXiv | `arxiv-pp-cli` on PATH (auto-installed during first-run setup via `npx -y @mvanhorn/[email protected] install arxiv --cli-only`) | always on if `arxiv-pp-cli` on PATH; fires on research/technical topics and stays quiet otherwise (relevance + 365-day recency gating) | yes (free, keyless) |
| Techmeme | `techmeme-pp-cli` on PATH (auto-installed via `... install techmeme --cli-only`) | always on if `techmeme-pp-cli` on PATH; searches Techmeme's live archive and keeps only headlines dated within the research window (undated headlines flow through as low-confidence) | yes (free, keyless) |
| Trustpilot | `trustpilot-pp-cli` on PATH (NOT auto-installed; install on demand via `npx -y @mvanhorn/[email protected] install trustpilot --cli-only`) + (`INCLUDE_SOURCES` contains `trustpilot`, `--search` includes `trustpilot`, **or** an explicit `--trustpilot-domain` / plan-level `trustpilot_domain`) | **opt-in, off by default**; company TrustScore, review count, and an AI summary, rather than individual reviews restricted to the research window. `--trustpilot-domain=<domain>` (and per-entity `trustpilot_domain` in `--competitors-plan`) auto-activates the source for that run and bypasses the brand-shape gate. `EXCLUDE_SOURCES=trustpilot` still wins. Bare company names resolve to the review-page domain through the CLI only when the source is active. A stale session triggers a one-time headless-Chrome WAF-cookie harvest. `LAST30DAYS_TRUSTPILOT_NO_BROWSER=1` skips the entire source, even with a warm session | yes (no API key; cookie-replay after the one-time harvest) |
diff --git a/README.de.md b/README.de.md
index b335b029..9abfb7c3 100644
--- a/README.de.md
+++ b/README.de.md
@@ -76,7 +76,7 @@ Wenn du dich mit einem CEO triffst: Hast du alle Tweets und YouTube-Transkripte
| **Hacker News** | Der Konsens der Entwickler. 825 Punkte, 899 Kommentare. Wo technische Leute wirklich streiten. |
| **Polymarket** | Keine Meinungen. Quoten. Gedeckt durch echtes Geld. 96 % Wahrscheinlichkeit bei Albumverkäufen. 4 % bei einer Übernahme. |
| **GitHub** | Für Personen: PR-Tempo, Top-Repos nach Sternen, Release Notes. Für Themen: Issues und Discussions. |
-| **Digg** | Kuratierte Story-Cluster aus Diggs AI-1000-Leaderboard (rund 1000 KI-Accounts mit hohem Signal auf X), mit zuordenbaren Inline-Zitaten und ganz ohne X-Authentifizierung. Wird automatisch aktiv, sobald `digg-pp-cli` im PATH liegt. |
+| **Digg** | Story-Cluster aus Diggs Rubriken AI und Tech, mit zuordenbaren Inline-Zitaten von X und ganz ohne X-Authentifizierung. Wird automatisch aktiv, sobald `digg-pp-cli` im PATH liegt. |
| **arXiv** | Die Fachartikel hinter dem Hype. Neue Forschung im Zeitfenster, kostenlos, ohne API-Schlüssel. Wird automatisch aktiv, sobald `arxiv-pp-cli` im PATH liegt (das Erst-Setup installiert es). |
| **Techmeme** | Die redaktionelle Ebene der Tech-News, begrenzt auf dein 30-Tage-Fenster. Kostenlos, ohne API-Schlüssel. Wird automatisch aktiv, sobald `techmeme-pp-cli` im PATH liegt (das Erst-Setup installiert es). |
| **LinkedIn** | Das berufliche Signal. Beiträge und Artikel, wobei Artikel als starkes Signal gewichtet werden. |
@@ -109,7 +109,7 @@ Die Synthese sortiert nach dem, womit echte Menschen tatsächlich interagiert ha
**Um Hiring-Signale zu lesen.** `/last30days Listen Labs --hiring-signals` – aktuelle Stellenanzeigen und Karriereseiten werden zu zitierten Belegen für Schwerpunktverschiebungen: Einstellungen in Enterprise Security, Customer Success, Infrastruktur oder Produktausbau. Der Bericht sagt, was das Hiring zu signalisieren scheint, nicht was die Roadmap liefern wird.
-**Um ein Thema vor seinem Höhepunkt zu finden.** Frag `/last30days what's exploding in AI agents?`, und die Skill wechselt in den Discovery-Modus: Die Engine durchkämmt Reddit-Kategorielisten, die Front- und Best-Stories von Hacker News, Diggs AI-1000-Feed und X, sofern du authentifiziert bist. Dein Agent bewertet die Vorschläge (Namen, Müllfilterung, inhaltliche Relevanz) und schreibt Podcast- und X-Artikel-Ansätze. Am Ende bekommst du 5 bis 10 nach Velocity sortierte Themen. Jedes Ergebnis enthält quellenübergreifende Zahlen, ein Momentum-Label und einen startklaren Folgebefehl `/last30days "<topic>"`.
+**Um ein Thema vor seinem Höhepunkt zu finden.** Frag `/last30days what's exploding in AI agents?`, und die Skill wechselt in den Discovery-Modus: Die Engine durchkämmt Reddit-Kategorielisten, die Front- und Best-Stories von Hacker News, Diggs AI- und Tech-Feeds und X, sofern du authentifiziert bist. Dein Agent bewertet die Vorschläge (Namen, Müllfilterung, inhaltliche Relevanz) und schreibt Podcast- und X-Artikel-Ansätze. Am Ende bekommst du 5 bis 10 nach Velocity sortierte Themen. Jedes Ergebnis enthält quellenübergreifende Zahlen, ein Momentum-Label und einen startklaren Folgebefehl `/last30days "<topic>"`.
**Wenn etwas erscheint.** `/last30days Kanye West` – Großbritannien hat sein Visum blockiert, das Wireless Festival wurde abgesagt, die Sponsoren sind abgesprungen. Aber BULLY stieg auf Platz 2 der Billboard-Charts ein. Fantano kam aus seinem „Yay sabbatical“ zurück, um es zu rezensieren (653.000 Aufrufe). Beim SoFi Homecoming holte er Lauryn Hill und Travis Scott für 44 Songs auf die Bühne. Polymarket: „Wird Kanye wieder twittern?“ 86 % Ja. 23 Reddit-Threads, 17 YouTube-Videos, 86.000 Upvotes.
diff --git a/README.es.md b/README.es.md
index a44b05c7..49603301 100644
--- a/README.es.md
+++ b/README.es.md
@@ -76,7 +76,7 @@ Si te vas a reunir con un CEO, ¿te has leído todos sus tuits y todas sus trans
| **Hacker News** | El consenso de los desarrolladores. 825 puntos, 899 comentarios. Donde la gente técnica discute de verdad. |
| **Polymarket** | No son opiniones. Son cuotas. Respaldadas por dinero real. 96 % de probabilidad en ventas de un álbum. 4 % en una adquisición. |
| **GitHub** | Para personas: ritmo de PR, mejores repositorios por estrellas, notas de versión. Para temas: issues y discusiones. |
-| **Digg** | Grupos de noticias seleccionados del ranking AI 1000 de Digg (unas 1000 cuentas de IA con mucha señal en X), con citas atribuibles integradas y sin necesidad de autenticarte en X. Se activa solo cuando `digg-pp-cli` está en el PATH. |
+| **Digg** | Grupos de noticias de las secciones AI y Tech de Digg, con citas atribuibles de X integradas y sin necesidad de autenticarte en X. Se activa solo cuando `digg-pp-cli` está en el PATH. |
| **arXiv** | Los artículos científicos que hay detrás del ruido. Investigación nueva dentro de la ventana, gratis y sin clave de API. Se activa solo cuando `arxiv-pp-cli` está en el PATH (la configuración inicial lo instala). |
| **Techmeme** | La capa editorial de la actualidad tecnológica, acotada a tu ventana de 30 días. Gratis y sin clave de API. Se activa solo cuando `techmeme-pp-cli` está en el PATH (la configuración inicial lo instala). |
| **LinkedIn** | La señal profesional. Publicaciones y artículos, con los artículos ponderados como señal fuerte. |
@@ -109,7 +109,7 @@ La síntesis ordena según aquello con lo que la gente real ha interactuado de v
**Para leer señales de contratación.** `/last30days Listen Labs --hiring-signals` — las ofertas de empleo y las páginas de carreras actuales se convierten en pruebas citadas de un cambio de prioridades: contratación en seguridad para empresa, customer success, infraestructura o expansión de producto. El informe dice lo que la contratación parece señalar, no lo que la hoja de ruta va a entregar.
-**Para encontrar el tema antes de su pico.** Pregunta `/last30days what's exploding in AI agents?` y la skill cambia a modo descubrimiento: el motor barre los listados por categoría de Reddit, la portada y las mejores historias de Hacker News, el feed AI 1000 de Digg y X si estás autenticado; tu agente evalúa las candidaturas (nombres, filtrado de ruido, interés real) y escribe enfoques para pódcast o para un artículo en X; después obtienes entre 5 y 10 temas ordenados por velocidad. Cada resultado incluye cifras de varias fuentes, una etiqueta de impulso y un comando `/last30days "<topic>"` listo para lanzar.
+**Para encontrar el tema antes de su pico.** Pregunta `/last30days what's exploding in AI agents?` y la skill cambia a modo descubrimiento: el motor barre los listados por categoría de Reddit, la portada y las mejores historias de Hacker News, los feeds AI y Tech de Digg y X si estás autenticado; tu agente evalúa las candidaturas (nombres, filtrado de ruido, interés real) y escribe enfoques para pódcast o para un artículo en X; después obtienes entre 5 y 10 temas ordenados por velocidad. Cada resultado incluye cifras de varias fuentes, una etiqueta de impulso y un comando `/last30days "<topic>"` listo para lanzar.
**Cuando sale algo nuevo.** `/last30days Kanye West` — el Reino Unido le bloqueó el visado, el Wireless Festival se canceló, los patrocinadores huyeron. Pero BULLY debutó en el número 2 del Billboard. Fantano volvió de su «Yay sabbatical» para reseñarlo (653.000 visualizaciones). En el SoFi Homecoming sacó al escenario a Lauryn Hill y a Travis Scott para 44 canciones. Polymarket: «¿Volverá Kanye a tuitear?» 86 % sí. 23 hilos de Reddit, 17 vídeos de YouTube, 86.000 votos positivos.
diff --git a/README.fr.md b/README.fr.md
index 97aac429..a4d77250 100644
--- a/README.fr.md
+++ b/README.fr.md
@@ -76,7 +76,7 @@ Si vous rencontrez un PDG, avez-vous lu tous ses tweets et toutes ses transcript
| **Hacker News** | Le consensus des développeurs. 825 points, 899 commentaires. Là où les gens techniques débattent vraiment. |
| **Polymarket** | Pas des opinions. Des cotes. Adossées à de l'argent réel. 96 % de probabilité sur des ventes d'album. 4 % sur une acquisition. |
| **GitHub** | Pour les personnes : rythme des PR, meilleurs dépôts par étoiles, notes de version. Pour les sujets : issues et discussions. |
-| **Digg** | Des groupes d'articles sélectionnés depuis le classement AI 1000 de Digg (environ 1000 comptes IA à fort signal sur X), avec des citations attribuables intégrées (sans authentification X). Activé automatiquement quand `digg-pp-cli` est présent dans le PATH. |
+| **Digg** | Des groupes d'articles issus des sections AI et Tech de Digg, avec des citations attribuables de X intégrées (sans authentification X). Activé automatiquement quand `digg-pp-cli` est présent dans le PATH. |
| **arXiv** | Les articles scientifiques derrière le battage médiatique. La recherche publiée dans la fenêtre, gratuit, sans clé API. Activé automatiquement quand `arxiv-pp-cli` est présent dans le PATH (la configuration initiale l'installe). |
| **Techmeme** | La couche éditoriale de l'actu tech, restreinte à votre fenêtre de 30 jours. Gratuit, sans clé API. Activé automatiquement quand `techmeme-pp-cli` est présent dans le PATH (la configuration initiale l'installe). |
| **LinkedIn** | Le signal professionnel. Posts et articles, les articles étant pondérés comme signal fort. |
@@ -109,7 +109,7 @@ La synthèse classe selon ce avec quoi de vraies personnes ont vraiment interagi
**Pour lire les signaux de recrutement.** `/last30days Listen Labs --hiring-signals` — les offres d'emploi et les pages carrières actuelles deviennent des preuves citées de changements de priorités : recrutements en sécurité entreprise, customer success, infrastructure ou expansion produit. Le rapport dit ce que le recrutement semble signaler, pas ce que la roadmap va livrer.
-**Pour repérer un sujet avant son pic.** Demandez `/last30days what's exploding in AI agents?` et la skill bascule en mode découverte : le moteur balaie les listings de catégories Reddit, la une et les meilleures histoires de Hacker News, le flux AI 1000 de Digg, et X si vous êtes authentifié ; votre agent évalue les candidats (noms, filtrage du bruit, intérêt réel) et rédige des angles pour un podcast ou un article X ; vous obtenez ensuite 5 à 10 sujets classés par vélocité. Chaque résultat comprend des chiffres multi-sources, une étiquette de momentum et une commande `/last30days "<topic>"` prête à lancer.
+**Pour repérer un sujet avant son pic.** Demandez `/last30days what's exploding in AI agents?` et la skill bascule en mode découverte : le moteur balaie les listings de catégories Reddit, la une et les meilleures histoires de Hacker News, les flux AI et Tech de Digg, et X si vous êtes authentifié ; votre agent évalue les candidats (noms, filtrage du bruit, intérêt réel) et rédige des angles pour un podcast ou un article X ; vous obtenez ensuite 5 à 10 sujets classés par vélocité. Chaque résultat comprend des chiffres multi-sources, une étiquette de momentum et une commande `/last30days "<topic>"` prête à lancer.
**Quand quelque chose sort.** `/last30days Kanye West` — le Royaume-Uni a bloqué son visa, le Wireless Festival est annulé, les sponsors ont fui. Mais BULLY est entré n° 2 au Billboard. Fantano est revenu de son « Yay sabbatical » pour le chroniquer (653 000 vues). SoFi Homecoming a fait monter Lauryn Hill et Travis Scott sur scène pour 44 titres. Polymarket : « Kanye tweetera-t-il de nouveau ? » 86 % de oui. 23 fils Reddit, 17 vidéos YouTube, 86 000 upvotes.
diff --git a/README.ja.md b/README.ja.md
index 22c10425..aad8406d 100644
--- a/README.ja.md
+++ b/README.ja.md
@@ -76,7 +76,7 @@ CEOと会うとして、直近30日間のツイートと YouTube の文字起こ
| **Hacker News** | 開発者の総意。825ポイント、899コメント。技術寄りの人たちが本気で議論している場所です。 |
| **Polymarket** | 意見ではなく、オッズ。実際のお金が裏付けています。アルバムの売上に96%、買収に4%といった具合です。 |
| **GitHub** | 人物について: PRの勢い、スター数の多いリポジトリ、リリースノート。トピックについて: Issue と Discussion。 |
-| **Digg** | Digg の AI 1000 リーダーボード(X 上でシグナルの強いAI関連アカウント約1000件)から集めたストーリークラスター。出典をたどれるインライン引用付きで、X の認証は不要です。`digg-pp-cli` が PATH にあると自動的に有効になります。 |
+| **Digg** | Digg の AI と Tech セクションから集めたストーリークラスター。X の投稿を出典をたどれるインライン引用として付け、X の認証は不要です。`digg-pp-cli` が PATH にあると自動的に有効になります。 |
| **arXiv** | 話題の裏側にある論文。対象期間に出た新しい研究が、無料・APIキーなしで手に入ります。`arxiv-pp-cli` が PATH にあると自動的に有効になります(初回セットアップでインストールされます)。 |
| **Techmeme** | テックニュースの編集レイヤーを、対象の30日間に絞って取得します。無料・APIキーなし。`techmeme-pp-cli` が PATH にあると自動的に有効になります(初回セットアップでインストールされます)。 |
| **LinkedIn** | ビジネス面のシグナル。投稿と記事を拾い、記事は強いシグナルとして重み付けします。 |
@@ -109,7 +109,7 @@ CEOと会うとして、直近30日間のツイートと YouTube の文字起こ
**採用シグナルを読むために。** `/last30days Listen Labs --hiring-signals` — 現在の求人ページやキャリアページが、注力領域の変化を示す引用可能な根拠になります。エンタープライズ向けセキュリティ、カスタマーサクセス、インフラ、プロダクト拡張といった採用の動きです。レポートが述べるのは「採用が何を示唆しているように見えるか」であって、「ロードマップが何を出すか」ではありません。
-**ピークを迎える前の話題を見つけるために。** `/last30days what's exploding in AI agents?` と尋ねると、スキルはディスカバリーモードに切り替わります。エンジンが Reddit のカテゴリー一覧、Hacker News のフロントページとベストストーリー、Digg の AI 1000 フィード、そして認証済みであれば X を横断してさらいます。次にエージェントが候補を審査し(名前の妥当性、ノイズの除去、記事になるか)、ポッドキャストや X 記事の切り口を書きます。最後に、勢いの強さで並べた5〜10件のトピックが返ってきます。各結果には情報源をまたいだ数値、モメンタムのラベル、そしてそのまま実行できる `/last30days "<topic>"` が付いてきます。
+**ピークを迎える前の話題を見つけるために。** `/last30days what's exploding in AI agents?` と尋ねると、スキルはディスカバリーモードに切り替わります。エンジンが Reddit のカテゴリー一覧、Hacker News のフロントページとベストストーリー、Digg の AI と Tech のフィード、そして認証済みであれば X を横断してさらいます。次にエージェントが候補を審査し(名前の妥当性、ノイズの除去、記事になるか)、ポッドキャストや X 記事の切り口を書きます。最後に、勢いの強さで並べた5〜10件のトピックが返ってきます。各結果には情報源をまたいだ数値、モメンタムのラベル、そしてそのまま実行できる `/last30days "<topic>"` が付いてきます。
**何かが出たとき。** `/last30days Kanye West` — イギリスがビザを却下、Wireless Festival は中止、スポンサーは離脱。それでも BULLY は Billboard 初登場2位。Fantano は「Yay sabbatical」から復帰してレビューを公開(65.3万回再生)。SoFi Homecoming では Lauryn Hill と Travis Scott を迎えて44曲を披露。Polymarket では「Kanye はまたツイートするか?」が「はい」86%。Reddit のスレッド23件、YouTube の動画17本、アップボート8.6万件。
diff --git a/README.md b/README.md
index 752c3e42..2aa76955 100644
--- a/README.md
+++ b/README.md
@@ -76,7 +76,7 @@ If you're meeting with a CEO, have you read all their tweets and YouTube transcr
| **Hacker News** | The developer consensus. 825 points, 899 comments. Where technical people actually argue. |
| **Polymarket** | Not opinions. Odds. Backed by real money. 96% confidence on album sales. 4% on an acquisition. |
| **GitHub** | For people: PR velocity, top repos by stars, release notes. For topics: issues and discussions. |
-| **Digg** | Curated story clusters from Digg's AI 1000 leaderboard (~1000 high-signal AI accounts on X), with attributable inline quotes (no X auth required). Auto-enabled when `digg-pp-cli` is on PATH. |
+| **Digg** | Story clusters from Digg's AI and Tech sections, with attributable inline quotes from X (no X auth required). Auto-enabled when `digg-pp-cli` is on PATH. |
| **arXiv** | The papers behind the hype. New research in the window, free, no API key. Auto-enabled when `arxiv-pp-cli` is on PATH (first-run setup installs it). |
| **Techmeme** | The tech-news editorial layer, date-windowed to your 30 days. Free, no API key. Auto-enabled when `techmeme-pp-cli` is on PATH (first-run setup installs it). |
| **LinkedIn** | The professional signal. Posts and articles, with articles weighted as high signal. |
@@ -109,7 +109,7 @@ The synthesis ranks by what real people actually engaged with. Social relevancy,
**To read hiring signals.** `/last30days Listen Labs --hiring-signals` - current jobs and careers pages become cited evidence for focus shifts: hiring into enterprise security, customer success, infrastructure, or product expansion. The report says what the hiring appears to signal, not what the roadmap will ship.
-**To find the topic before it peaks.** Ask `/last30days what's exploding in AI agents?` and the skill switches to discovery mode: the engine sweeps Reddit category listings, Hacker News front/best stories, Digg's AI 1000 feed, and X when authenticated; your agent judges the nominations (names, junk filtering, content-worthiness) and writes podcast / X-article angles; then you get 5-10 velocity-ranked topics. Every result includes cross-source numbers, a momentum label, and a ready-to-run `/last30days "<topic>"` follow-up.
+**To find the topic before it peaks.** Ask `/last30days what's exploding in AI agents?` and the skill switches to discovery mode: the engine sweeps Reddit category listings, Hacker News front/best stories, Digg's AI and Tech feeds, and X when authenticated; your agent judges the nominations (names, junk filtering, content-worthiness) and writes podcast / X-article angles; then you get 5-10 velocity-ranked topics. Every result includes cross-source numbers, a momentum label, and a ready-to-run `/last30days "<topic>"` follow-up.
**When something drops.** `/last30days Kanye West` - UK blocked his visa, Wireless Festival canceled, sponsors fled. But BULLY debuted #2 on Billboard. Fantano came back from his "Yay sabbatical" to review it (653K views). SoFi Homecoming brought out Lauryn Hill and Travis Scott for 44 songs. Polymarket: "Will Kanye tweet again?" 86% Yes. 23 Reddit threads, 17 YouTube videos, 86K upvotes.
diff --git a/README.pt-BR.md b/README.pt-BR.md
index be1e7e11..e4fea4cb 100644
--- a/README.pt-BR.md
+++ b/README.pt-BR.md
@@ -76,7 +76,7 @@ Se você vai se reunir com um CEO, já leu todos os tweets e todas as transcriç
| **Hacker News** | O consenso da turma de desenvolvimento. 825 pontos, 899 comentários. Onde o pessoal técnico discute de verdade. |
| **Polymarket** | Não são opiniões. São probabilidades. Lastreadas em dinheiro de verdade. 96 % de confiança em vendas de um álbum. 4 % em uma aquisição. |
| **GitHub** | Para pessoas: ritmo de PRs, principais repositórios por estrelas, notas de versão. Para assuntos: issues e discussions. |
-| **Digg** | Agrupamentos de histórias curados a partir do ranking AI 1000 do Digg (cerca de 1000 contas de IA com alto sinal no X), com citações atribuíveis embutidas e sem exigir autenticação no X. Ativa sozinho quando `digg-pp-cli` está no PATH. |
+| **Digg** | Agrupamentos de histórias das seções AI e Tech do Digg, com citações atribuíveis do X embutidas e sem exigir autenticação no X. Ativa sozinho quando `digg-pp-cli` está no PATH. |
| **arXiv** | Os artigos científicos por trás do hype. Pesquisa nova dentro da janela, de graça e sem chave de API. Ativa sozinho quando `arxiv-pp-cli` está no PATH (a configuração inicial instala). |
| **Techmeme** | A camada editorial do noticiário de tecnologia, limitada à sua janela de 30 dias. De graça e sem chave de API. Ativa sozinho quando `techmeme-pp-cli` está no PATH (a configuração inicial instala). |
| **LinkedIn** | O sinal profissional. Publicações e artigos, com os artigos ponderados como sinal forte. |
@@ -109,7 +109,7 @@ A síntese ordena pelo que as pessoas de verdade realmente engajaram. Relevânci
**Para ler sinais de contratação.** `/last30days Listen Labs --hiring-signals` — as vagas e páginas de carreira atuais viram evidência citada de mudança de prioridade: contratação em segurança para empresas, customer success, infraestrutura ou expansão de produto. O relatório diz o que a contratação parece sinalizar, não o que o roadmap vai entregar.
-**Para achar o assunto antes do pico.** Pergunte `/last30days what's exploding in AI agents?` e a skill muda para o modo descoberta: o motor varre as listagens por categoria do Reddit, a capa e as melhores histórias do Hacker News, o feed AI 1000 do Digg e o X quando você está autenticado; seu agente avalia as indicações (nomes, filtragem de ruído, se rende conteúdo) e escreve ângulos para podcast ou para um artigo no X; depois você recebe de 5 a 10 assuntos ordenados por velocidade. Cada resultado traz números de várias fontes, um rótulo de momentum e um comando `/last30days "<topic>"` pronto para rodar.
+**Para achar o assunto antes do pico.** Pergunte `/last30days what's exploding in AI agents?` e a skill muda para o modo descoberta: o motor varre as listagens por categoria do Reddit, a capa e as melhores histórias do Hacker News, os feeds AI e Tech do Digg e o X quando você está autenticado; seu agente avalia as indicações (nomes, filtragem de ruído, se rende conteúdo) e escreve ângulos para podcast ou para um artigo no X; depois você recebe de 5 a 10 assuntos ordenados por velocidade. Cada resultado traz números de várias fontes, um rótulo de momentum e um comando `/last30days "<topic>"` pronto para rodar.
**Quando alguma coisa é lançada.** `/last30days Kanye West` — o Reino Unido bloqueou o visto dele, o Wireless Festival foi cancelado, os patrocinadores fugiram. Mas BULLY estreou em 2º na Billboard. Fantano voltou do "Yay sabbatical" para resenhar o disco (653 mil visualizações). No SoFi Homecoming, ele levou Lauryn Hill e Travis Scott ao palco para 44 músicas. Polymarket: "Kanye vai tuitar de novo?" 86 % sim. 23 threads no Reddit, 17 vídeos no YouTube, 86 mil votos positivos.
diff --git a/README.zh-CN.md b/README.zh-CN.md
index 68ff385a..683ab9fd 100644
--- a/README.zh-CN.md
+++ b/README.zh-CN.md
@@ -79,7 +79,7 @@ Google 聚合编辑选出的内容,`/last30days` 搜索真实的人。
| **Hacker News** | 开发者共识:825 分、899 条评论,技术从业者真正交锋的地方。 |
| **Polymarket** | 不是观点,而是由真金白银支撑的概率:专辑销量 96%,收购概率 4%。 |
| **GitHub** | 搜人时查看 PR 速度、按 Star 排名的热门仓库和发行说明;搜主题时查看 Issue 与 Discussion。 |
-| **Digg** | 来自 Digg AI 1000 排行榜(约 1,000 个高信号 X 账号)的精选话题聚类,包含可追溯的行内引用,无需 X 认证。当 PATH 中存在 `digg-pp-cli` 时自动启用。 |
+| **Digg** | 来自 Digg AI 与 Tech 板块的话题聚类,包含可追溯的 X 行内引用,无需 X 认证。当 PATH 中存在 `digg-pp-cli` 时自动启用。 |
| **arXiv** | 热点背后的论文。免费查找时间窗口内的新研究,无需 API 密钥。当 PATH 中存在 `arxiv-pp-cli` 时自动启用(首次配置会安装)。 |
| **Techmeme** | 科技新闻的编辑视角,并按你设定的 30 天窗口筛选。免费且无需 API 密钥。当 PATH 中存在 `techmeme-pp-cli` 时自动启用(首次配置会安装)。 |
| **LinkedIn** | 职业领域的信号。搜索帖子和文章,其中文章被视为高价值信号。 |
@@ -112,7 +112,7 @@ Google 聚合编辑选出的内容,`/last30days` 搜索真实的人。
**判断招聘信号。** `/last30days Listen Labs --hiring-signals`——把最新职位和招聘页面变成有引用依据的证据,从中判断公司是否正转向企业安全、客户成功、基础设施或产品扩张。报告只解释招聘看起来释放了什么信号,不会武断预测路线图一定会交付什么。
-**在话题爆发前发现它。** 输入 `/last30days what's exploding in AI agents?`,Skill 会切换到发现模式:引擎扫描 Reddit 分类列表、Hacker News 的 front/best 故事、Digg AI 1000 信息流,以及认证后的 X;随后由你的智能体评审候选主题(命名、过滤垃圾、判断内容价值),并写出播客或 X 长文的切入角度;最终给出 5–10 个按增长速度排序的话题。每条结果都包含跨平台数据、势头标签,以及可直接运行的 `/last30days "<topic>"` 后续命令。
+**在话题爆发前发现它。** 输入 `/last30days what's exploding in AI agents?`,Skill 会切换到发现模式:引擎扫描 Reddit 分类列表、Hacker News 的 front/best 故事、Digg AI 与 Tech 信息流,以及认证后的 X;随后由你的智能体评审候选主题(命名、过滤垃圾、判断内容价值),并写出播客或 X 长文的切入角度;最终给出 5–10 个按增长速度排序的话题。每条结果都包含跨平台数据、势头标签,以及可直接运行的 `/last30days "<topic>"` 后续命令。
**突发事件发生时。** `/last30days Kanye West`——英国拒绝其签证,Wireless Festival 取消演出,赞助商纷纷离场;但《BULLY》首周登上 Billboard 第二名。Fantano 结束自己的 “Yay sabbatical” 回归评测(65.3 万次观看);SoFi Homecoming 请来 Lauryn Hill 和 Travis Scott,共演出 44 首歌。Polymarket:“Kanye 还会再发推吗?”86% 认为会。共找到 23 个 Reddit 主题、17 个 YouTube 视频和 8.6 万次赞同。
diff --git a/changelog.d/+digg-ai-tech-sections.fixed.md b/changelog.d/+digg-ai-tech-sections.fixed.md
new file mode 100644
index 00000000..cf94a516
--- /dev/null
+++ b/changelog.d/+digg-ai-tech-sections.fixed.md
@@ -0,0 +1 @@
+Digg results work again with digg-pp-cli 2026.10.1, which rebuilt search after Digg removed its search API. last30days now searches Digg's AI and Tech sections, keeps an empty answer from a freshly synced copy as "no matches" instead of an error, and marks Digg partial in the brief, with the `digg-pp-cli sync --topics ai,tech` command to run, when the CLI searched only its newest stories.
diff --git a/skills/last30days/scripts/lib/digg.py b/skills/last30days/scripts/lib/digg.py
index 4697ea4d..2ac3a8dc 100644
--- a/skills/last30days/scripts/lib/digg.py
+++ b/skills/last30days/scripts/lib/digg.py
@@ -1,16 +1,20 @@
-"""Digg AI 1000 source for last30days.
+"""Digg source for last30days.
Shells out to ``digg-pp-cli`` (read-only, no auth required) to surface
-clustered stories curated from ~1000 high-signal AI accounts on X. Each
-cluster carries a published TLDR, a curatorial rank, and a list of X
-posts that can be fetched as inline quotes.
+clustered stories from Digg's AI and Tech sections. Each cluster carries
+a published TLDR, a section rank, and a list of X posts that can be
+fetched as inline quotes.
Activation gate: this source is only available when ``digg-pp-cli`` is
on PATH. ``pipeline.available_sources`` checks ``shutil.which`` before
including ``digg`` in the source list. The functions below also detect
the missing-binary case as a defensive fallback.
-Primary path: ``digg-pp-cli search <topic> --since 30d --agent --limit N``.
+Primary path: ``digg-pp-cli search <topic> --topic ai,tech --since 30d
+--agent --limit N``. The CLI searches a local copy of each section's feed
+that ``digg-pp-cli sync`` builds. When a section has no recent sync it
+searches only the newest live pages and sets ``meta.stale``; see
+``stale_note``.
Optional enrichment: ``digg-pp-cli posts <clusterUrlId> --agent --by rank
--limit M`` for the top K clusters in default/deep depth, attaching the
top-ranked X posts to each cluster's ``posts`` field.
@@ -31,6 +35,17 @@ from .relevance import token_overlap_relevance
CLI_BIN = "digg-pp-cli"
+# Digg sections searched. Digg also covers world-business, sports, science,
+# gaming and entertainment; last30days keeps its Digg lane on AI and tech.
+SECTIONS = "ai,tech"
+
+# Outcome note for a search answered from a stale local copy. Kept under the
+# 140-character outcome note limit in render.
+STALE_NOTE = (
+ "no recent local sync, so only the newest stories were searched; "
+ f"run `{CLI_BIN} sync --topics {SECTIONS}` for full coverage"
+)
+
# Per-depth knobs.
DEPTH_CONFIG = {
"quick": 8,
@@ -50,6 +65,8 @@ ENRICH_CONFIG = {
# Reddit/HN/YouTube/TikTok/GitHub enrichment.
POSTS_PER_CLUSTER = 5
+# A stale section makes the CLI fetch up to 8 live feed pages per section
+# before answering (about 8s for ai,tech), well inside this budget.
SEARCH_TIMEOUT = 30
POSTS_TIMEOUT = 15
@@ -113,6 +130,8 @@ def _build_search_args(query: str, limit: int) -> List[str]:
CLI_BIN,
"search",
query,
+ "--topic",
+ SECTIONS,
"--since",
"30d",
"--agent",
@@ -167,12 +186,18 @@ def _empty_local_fallback_error(data: Dict[str, Any], stderr: str = "") -> Optio
reaches the user as ``Digg: 0 items (no results)`` instead of an error.
A non-empty local store is real data and is left alone.
+ CLI builds from 2026.10.1 always send ``meta.stale`` and exit non-zero
+ themselves when they have no usable data, so an empty local answer from
+ them is a real "no matches" and is left alone too.
+
The CLI reports the live failure only on stderr; ``meta.reason`` is
preferred if a future CLI release adds it to the envelope.
"""
meta = data.get("meta")
if not isinstance(meta, dict) or meta.get("source") != "local":
return None
+ if "stale" in meta:
+ return None
if data.get("results"):
return None
reason = str(meta.get("reason") or "").strip()[:_FALLBACK_REASON_MAX]
@@ -240,7 +265,7 @@ def search_digg(
to_date: str,
depth: str = "default",
) -> Dict[str, Any]:
- """Search Digg AI 1000 clusters via digg-pp-cli.
+ """Search Digg AI and Tech clusters via digg-pp-cli.
Args:
topic: search query.
@@ -256,7 +281,7 @@ def search_digg(
if not topic or not topic.strip():
return {"results": []}
cmd = _build_search_args(topic, limit)
- _log(f"search '{topic}' (limit={limit}, since=30d)")
+ _log(f"search '{topic}' (sections={SECTIONS}, limit={limit}, since=30d)")
response, stderr = _invoke_cli(cmd, timeout=SEARCH_TIMEOUT)
fallback_error = _empty_local_fallback_error(response, stderr)
if fallback_error:
@@ -264,9 +289,30 @@ def search_digg(
return {"results": [], "error": fallback_error}
n = len(response.get("results") or [])
_log(f"found {n} clusters")
+ note = stale_note(response)
+ if note:
+ _log(note)
return response
+def stale_note(response: Dict[str, Any]) -> Optional[str]:
+ """Describe a search answered from a stale local copy, else None.
+
+ ``digg-pp-cli`` sets ``meta.stale`` when a searched section has no
+ completed sync in the last 6 hours. It then searches only the newest
+ live feed pages, so older stories in the 30-day window can be missing.
+ """
+ meta = response.get("meta") if isinstance(response, dict) else None
+ if not isinstance(meta, dict) or meta.get("stale") is not True:
+ return None
+ return STALE_NOTE
+
+
+def is_stale_note(detail: Optional[str]) -> bool:
+ """True when an outcome detail carries the stale-copy note."""
+ return bool(detail) and STALE_NOTE in detail
+
+
def _build_url(cluster_url_id: str) -> str:
return f"https://di.gg/ai/{cluster_url_id}"
diff --git a/skills/last30days/scripts/lib/pipeline.py b/skills/last30days/scripts/lib/pipeline.py
index c6c2b2fd..abb3dcb1 100644
--- a/skills/last30days/scripts/lib/pipeline.py
+++ b/skills/last30days/scripts/lib/pipeline.py
@@ -674,7 +674,7 @@ def _fetch_discovery_source(
if source == "digg":
result = digg.search_digg(plan.domain, from_date, to_date, depth=depth)
items = digg.parse_digg_response(result, query=plan.domain)
- # Digg is an AI-focused broad listing, so keep only domain-bearing
+ # Digg's AI and Tech sections are a broad listing, so keep only domain-bearing
# clusters when scoped; global trending keeps the whole feed.
if keyword_gate:
items = [
@@ -4162,7 +4162,12 @@ def _finalize_source_status(
elif state == health.OK and not count:
state = outcome.lane_failure_state or schema.NO_RESULTS
elif state == schema.PARTIAL and not count:
- state = http.classify_failure(message=detail or "")
+ if source == "digg" and digg.is_stale_note(detail):
+ # A stale Digg copy that matched nothing is an empty answer
+ # with a coverage note, not a failed search.
+ state = schema.NO_RESULTS
+ else:
+ state = http.classify_failure(message=detail or "")
finalized[source] = schema.SourceOutcome(
source=source,
state=state,
@@ -6182,6 +6187,9 @@ def _retrieve_stream_impl(
# Enrichment with attached X posts is deferred to
# _finalize_items_by_source so it runs on the items that actually
# survive dedupe rather than on top-K of the raw fanout.
+ stale = digg.stale_note(result)
+ if stale:
+ return items, _outcome_artifact(health.PARTIAL, stale)
return items, _result_outcome_artifact(source, result)
if source == "arxiv":
result = arxiv.search_arxiv(subquery.search_query, from_date, to_date, depth=depth)
diff --git a/tests/test_digg.py b/tests/test_digg.py
index 397dce00..949a9b4e 100644
--- a/tests/test_digg.py
+++ b/tests/test_digg.py
@@ -267,6 +267,7 @@ def test_search_digg_passes_since_30d(monkeypatch):
assert captured["cmd"][captured["cmd"].index("--since") + 1] == "30d"
assert "--agent" in captured["cmd"]
assert captured["cmd"][:3] == [digg.CLI_BIN, "search", "openclaw"]
+ assert captured["cmd"][captured["cmd"].index("--topic") + 1] == "ai,tech"
def test_search_digg_subproc_timeout_returns_empty(monkeypatch):
@@ -616,3 +617,117 @@ class TestLiveDigg:
if __name__ == "__main__":
pytest.main([__file__, "-v"])
+
+
+# === digg-pp-cli 2026.10.1+ envelopes (local feed copy with meta.stale) ===
+
+
+def _feed_envelope(*, source: str, stale: bool, results: list) -> subproc.SubprocResult:
+ return _stdout_for(
+ {
+ "meta": {
+ "count": len(results),
+ "query": "openai",
+ "since": "30d",
+ "source": source,
+ "stale": stale,
+ "topics": ["ai", "tech"],
+ },
+ "results": results,
+ }
+ )
+
+
+def test_search_digg_fresh_local_empty_is_no_matches(monkeypatch):
+ """A synced store with nothing matching is a real empty answer, not a fallback."""
+ monkeypatch.setattr(digg.shutil, "which", lambda _: "/fake/path")
+ monkeypatch.setattr(
+ digg.subproc,
+ "run_with_timeout",
+ lambda *a, **k: _feed_envelope(source="local", stale=False, results=[]),
+ )
+ out = digg.search_digg("openai", "2026-04-09", "2026-05-09")
+ assert out["results"] == []
+ assert "error" not in out
+ assert digg.stale_note(out) is None
+
+
+def test_search_digg_stale_result_keeps_items_and_notes_sync(monkeypatch):
+ monkeypatch.setattr(digg.shutil, "which", lambda _: "/fake/path")
+ monkeypatch.setattr(
+ digg.subproc,
+ "run_with_timeout",
+ lambda *a, **k: _feed_envelope(
+ source="local+live", stale=True, results=[_cluster()]
+ ),
+ )
+ out = digg.search_digg("openai", "2026-04-09", "2026-05-09")
+ assert len(out["results"]) == 1
+ note = digg.stale_note(out)
+ assert note is not None
+ assert "digg-pp-cli sync --topics ai,tech" in note
+
+
+def test_stale_note_ignores_old_envelopes():
+ assert digg.stale_note({"results": []}) is None
+ assert digg.stale_note({"meta": {"source": "local"}, "results": []}) is None
+ assert digg.stale_note({"meta": {"stale": "true"}, "results": []}) is None
+
+
+def test_stale_search_marks_digg_partial_in_pipeline(monkeypatch):
+ from lib import health, pipeline, schema
+
+ monkeypatch.setattr(digg.shutil, "which", lambda _: "/fake/path")
+ monkeypatch.setattr(
+ digg.subproc,
+ "run_with_timeout",
+ lambda *a, **k: _feed_envelope(
+ source="local+live", stale=True, results=[_cluster()]
+ ),
+ )
+ subquery = schema.SubQuery(
+ label="primary",
+ search_query="openai",
+ ranking_query="openai",
+ sources=["digg"],
+ )
+ items, artifact = pipeline._retrieve_stream_impl(
+ topic="openai",
+ subquery=subquery,
+ source="digg",
+ config={},
+ depth="default",
+ date_range=("2026-04-09", "2026-05-09"),
+ runtime=None,
+ mock=False,
+ )
+ assert len(items) == 1
+ outcome = artifact["_source_outcome"]
+ assert outcome["state"] == health.PARTIAL
+ assert "sync --topics ai,tech" in outcome["detail"]
+
+
+def test_stale_partial_with_no_items_finalizes_as_no_results():
+ """A stale search that matched nothing keeps its sync note and is not an error."""
+ from lib import health, pipeline, schema
+
+ outcomes = {
+ "digg": schema.SourceOutcome(
+ source="digg",
+ state=health.PARTIAL,
+ items_returned=0,
+ attempted=True,
+ detail=digg.STALE_NOTE,
+ ),
+ "reddit": schema.SourceOutcome(
+ source="reddit",
+ state=health.PARTIAL,
+ items_returned=0,
+ attempted=True,
+ detail="HTTP 503 from upstream",
+ ),
+ }
+ final = pipeline._finalize_source_status(outcomes, {})
+ assert final["digg"].state == schema.NO_RESULTS
+ assert final["digg"].detail == digg.STALE_NOTE
+ assert final["reddit"].state != schema.NO_RESULTS