Compress a two-hour interview into ten minutes: these 8 podcast deep-read sites solve the 'can't finish listening' problem

There are too many worthwhile podcasts and not enough time.

A one-hour episode gets split across three days of commuting. A two-hour interview sits in your queue for two weeks. This is the reality for many people.

The tweet author Orange AI (@oran_ge) listed 8 podcast deep-read websites. Each episode is automatically transcribed into an article, so you can read the main content of a one-hour episode in ten minutes. These sites were built by Xiangyang (@vista8); the tweet calls him a 'wish-granting coding Agent'—meaning you make a request and he builds it.

What each of the eight sites covers:

• Zhang Xiaojun Business Interview Deep Reads (https://zhangxiaojun.qiaomu.ai): Deep reads of multi-hour long interviews, capturing first-hand thinking from AI, tech companies, and investors.
• 42 Chapters Deep Reads (https://42zhangjing.qiaomu.ai): Qu Kai talks with entrepreneurs and investors, drilling into one specific question until it is fully explored.
• LateTalk Deep Reads (https://latetalk.qiaomu.ai): First-hand tech interviews produced by LatePost, covering large models, embodied AI, and internet companies.
• Crossroads Deep Reads (https://crossing.qiaomu.ai): Finds and interviews active participants in the AI era, focusing on the changes AI brings to various industries.
• AI Alchemy Deep Reads (https://aialchemy.qiaomu.ai): Xu Wenhao and Ren Xin discuss the fundamental changes and business opportunities brought by AI.
• 35Huan Deep Reads (https://35huan.qiaomu.ai): Liu Fei invites urban insiders to chat, discovering different perspectives and attitudes toward life.
• Zhixing Bistro Deep Reads (https://zhixing.qiaomu.ai): Produced by Youzhi Youxing, it talks about investing and personal finance, and also about how to live better.
• Mianji Deep Reads (https://mianji.qiaomu.ai): Each episode invites someone with frontline experience to explain an industry or profession clearly.

The key difference is that these sites produce fine-grained transcriptions, not summaries. A summary compresses the content into an outline and loses half the information. A transcription preserves the original wording, the line of argument, and specific numbers. After ten minutes of reading, you know what the episode covers and whether it is worth spending two hours listening to the original audio.

One commenter did the math: compressing one hour into ten minutes means the host's filler ratio is at least four-fifths. It is half a joke, but it points to a fact: most podcasts do not have a high density of useful information. The real value of deep-read sites is filtering—turning 'I can't finish listening' into 'first decide whether it's worth listening to.'

There are also several expanded approaches in the comments.

Commenter Zhao Er built an Agent workflow: every morning it fetches subscription updates, confirms which episodes to interpret, downloads the audio and uploads it to Google Drive, and uses a Gemini scheduled task to transcribe it—trying captions first, then falling back to audio transcription if captions are unavailable. It outputs three types of files: an in-depth analytical guide, a complete polished transcript, and professional interpretation notes.

Xiangyang Qiaomu added that Obsidian can install the Qiaomu AI RSS plugin, which already has these podcasts' RSS feeds built in. You can see all updates in one interface, highlight and take notes, and even ask AI questions. Plugin address: https://community.obsidian.md/plugins/qiaomu-ai-rss

Another commenter, Tony, made shengwenji.com, which specializes in illustrated verbatim transcripts. Address: https://shengwenji.com/explore

There are more and more tools, but the idea is consistent: first make long content shorter, then decide whether to spend time on it. As for the filler ratio, hosts may disagree, but the comment section has already finished doing the math.