As AI tools continue to grow more capable, I used Cursor to build another new project: a fully automated podcast called ‘Security Fresh News.’ It updates automatically at 7 every morning, covers the latest cybersecurity news from around the world, and runs through the entire workflow with almost no human intervention. Today I’ll talk about how I built it with AI and automation tools.
1. Why Build an Automated Podcast?
As a security professional, I follow a large amount of security news and vulnerability intelligence every day. Manually compiling, recording, editing, and publishing it would be time-consuming and tedious. The emergence of large AI models and automation tools made me wonder whether I could produce a podcast fully automatically. So I used Cursor and a collection of cloud services to turn the idea into reality.
2. What Does the Overall Architecture Look Like?
Data collection：The latest security news is fetched from multiple RSS sources on a daily schedule.
Article processing：Fetched news is first stored in an SQLite database, then summarized, simplified, and rewritten by an AI Agent (based on Alibaba Cloud Bailian DashScope) to produce content suitable for audio narration.
Speech synthesis：DashScope’s TTS (text-to-speech) service converts AI-generated content into audio files.
Podcast generation and publishing：A Podcast XML Feed is generated automatically and uploaded to object storage (OSS), then delivered to platforms such as Apple Podcasts and Xiaoyuzhou.
Scheduling and monitoring：A Scheduler automatically orchestrates every step, and sends alerts to my iPhone when something goes wrong.
3. AI Agents: The ‘Brain’ That Generates the Content
The AI Agent Layer is the core of the automated podcast system. It mainly consists of three modules:
News simplification(one agent application) This agent extracts useful information from RSS subscription content and writes it as a news item.
Script writing(one Alibaba Cloud agent orchestration application containing three agents) The Writer agent coherently combines the extracted and simplified news items into one broadcast script; the Reviewer checks the script for sensitive domestic information; and the Pronunciation agent corrects special terms, such as converting Pwn2Own to Pwn To Own, SQLI to SQL injection, and att&ck to attack.
Podcast text writing(two agent applications) These write the podcast episode title and its description (Show Notes), respectively.
AI agents understand the news’s key points, extract important information, and avoid mechanical concatenation. Review removes sensitive information, while the generated content remains coherent and natural for direct TTS synthesis. The style can be adjusted, for example to be more conversational or concise, and improved based on playback results.
4. Details of Full Automation
Data collection and processing require no human participation; AI automatically filters, summarizes, and polishes the news.
Speech synthesis uses high-quality TTS with a natural sound. It is somewhat mechanical but still listenable; the priority is getting the news quickly.
Publishing is fully automated: audio and the Feed are pushed to major platforms simultaneously, so listeners hear the latest security news every morning.
The entire system uses SQLite for storage; it is lightweight, efficient, and easy to maintain.
5. My Experience Making a Podcast with AI
AI and automation have greatly improved efficiency. Producing one video used to take several hours; now the system completes every step automatically on schedule. I only check the logs occasionally. Content quality is stable: AI-generated scripts are concise, and the TTS voice sounds natural.
There are many shortcomings. Some AI TTS pronunciations are inaccurate and cannot be corrected: The Hacker News is always pronounced The Hackers News, and numbers are sometimes read in English. Prompt engineering may correct these issues; I am still exploring.
Subscribe to My Podcast
If you are also interested in cybersecurity, please subscribe to my podcast ‘Security Fresh News’:
🍎Apple Podcast: https://podcasts.apple.com/us/podcast/安全新鲜事/id1818261420
📻Xiaoyuzhou: https://www.xiaoyuzhoufm.com/podcast/683e7a0938dcc57c6444f889
🔗RSS subscription: https://podcast.liyang.tech/feed.xml
Feel free to share ideas about new ways to use automation and AI in content creation! Click the original article link to open Xiaoyuzhou.