
Apache Airflow & Magpie Meetup 2026 – Taipei
About this event
(English version below) 活動介紹 歡迎參加 Apache Airflow & Magpie Meetup 2026 台北場! 這次我們邀請到 Jarek Potiuk,Apache Airflow PMC member,與大家分享 Apache Magpie,以及 AI 時代下開源專案維護者面臨的挑戰與機會。 除了技術分享,我們也希望藉由這次聚會,讓台灣的 Apache Airflow 使用者、開發者與開源貢獻者有機會交流經驗、認識彼此。 無論你是 Airflow 的長期使用者、開源貢獻者,或只是對工作流程編排(Workflow Orchestration)與開源技術感興趣,都非常歡迎參加! 📅 活動資訊 \- 日期: 2026 年 11 月 24 日(星期二) \- 時間: 18:30–20:30(UTC\+8) \- 地點: 台北市敦化南路二段 206 號 12 樓 \- 語言: 台灣華語/English 🎤 議程 Talks • Eight Times the Reports: Surviving the AI Vulnerability Flood at Apache Airflow Speaker: Jarek Potiuk — Apache Airflow PMC Member 2025 年,Apache Airflow 的安全團隊總共追蹤了約 30 份漏洞通報。然而到了 2026 年,通報數量已經突破 250 份,大約是前一年的八倍。光是夏季某個月份收到的通報,就超過了 2024 年全年的總數。 這些通報大多看起來都像是嚴重的安全漏洞:內容充滿自信、描述詳盡,而且經常附上 CVSS 評分。 但問題在於,這波通報並不全是無效資訊。雖然有不少通報並非真正的漏洞,或只是安全強化建議,但其中仍有超過 100 份最終成為 CVE,數量約為前一年的六倍。 AI 工具既能找出真正的漏洞,也會憑空捏造不存在的問題。 這不但沒有讓事情變得更簡單,反而更加棘手:我們不能忽略大量湧入的通報,卻也沒有足夠的人力逐一審查。 因此,我們決定用 AI 對抗 AI,打造了 Apache Magpie: • 從郵件列表匯入安全漏洞通報,並建立追蹤紀錄 • 自動建議漏洞分類 • 偵測重複通報,並協助分配 CVE 編號 • 持續同步追蹤紀錄、相關郵件討論串與修復用的 Pull Request • 協助草擬回覆漏洞通報者的郵件 每個步驟都必須經過安全團隊成員審核與批准。 這場演講將分享我們在第一線處理大量安全漏洞通報的實戰經驗:這波通報洪流究竟是什麼樣子、哪些工作適合自動化、哪些事情我們堅持交由人類決定,以及過程中犯過的錯誤與學到的教訓。 最後,也會介紹其他開源專案如何建立類似的防禦機制,在這波浪潮席捲而來之前做好準備。 • From N Consumers to One: How Airflow’s Triggerer Shares a Kafka Consumer Speaker: Po-An Yang — Apache Kafka Committer, Airflow Contributor Po-An 會介紹 Apache Airflow 3.3 的新功能 SharedStreamProducer,以及 trigger 如何共用 Kafka Consumer,降低針對同個 topic 的 Kafka 連線。 🕒 流程 \* 18:30–18:45 報到與交流 \* 18:45–19:30 Eight Times the Reports: Surviving the AI Vulnerability Flood at Apache Airflow \* 19:30–20:00 From N Consumers to One: How Airflow’s Triggerer Shares a Kafka Consumer \* 20:00–20:30 Q&A & Networking 💻 白天 Development Sprint 當天 10:00–17:00,我們也會在同一個場地舉辦 Apache Airflow Development Sprint 2026。 歡迎帶著筆電,一起參與 Apache Airflow 的開發、Review PR、改善文件,或與其他開源貢獻者交流。 白天與晚上的活動分開報名,可以自由選擇參加其中一場或兩場! 📍 場地與致謝 台北市敦化南路二段 206 號 12 樓。 特別感謝 TAIONE 開源基金會 (https://taione.org/),支持台灣開源社群的交流與協作。 \-\-\- About the Meetup Join us for the Apache Airflow & Magpie Meetup 2026 in Taipei! We're excited to welcome Jarek Potiuk, an Apache Airflow PMC member, who will introduce Apache Magpie and discuss the challenges and opportunities of maintaining open-source projects in the age of AI. This meetup is also an opportunity for Apache Airflow users, developers, and open-source contributors in Taiwan to share experiences, exchange ideas, and connect with the community. Whether you're an experienced Airflow user, a contributor, or simply curious about workflow orchestration and open source, everyone is welcome! 📅 Event Details \- Date: Tuesday\, November 24\, 2026 \- Time: 18:30–20:30 \(UTC\+8\) \- Venue: 12F\.\, No\. 206\, Sec\. 2\, Dunhua S\. Rd\.\, Taipei City\, Taiwan \- Language: Taiwanese Mandarin / English 🎤 Talks • Eight Times the Reports: Surviving the AI Vulnerability Flood at Apache Airflow Speaker: Jarek Potiuk — Apache Airflow PMC Member In 2025 Airflow's security team tracked about 30 reports. In 2026 we passed 250, roughly eight times as many, and in one summer month we got more than in all of 2024. Most of them read like a serious finding: confident, detailed, often with a CVSS score attached. The twist is that the flood isn't all junk. Plenty of reports were invalid or only hardening suggestions, but well over 100 became CVEs, around six times as many as the year before. AI tools find real bugs as well as imaginary ones. That makes it worse, not better: you can't ignore the flood, and you can't afford to read it all by hand. So we fought AI with AI. Apache Magpie: • imports reports from the mailing list into a tracker • suggests a classification • catches duplicates and helps allocate CVEs • keeps every tracker in sync with its email thread and fix PRs • drafts replies to reporters A human on the security team approves every step. This is the story from the trenches: what the flood looks like, what we automated and what we never will, the mistakes along the way, and how your project can se
Questions & comments
Ask the host anything — replies are visible to everyone.
—