Content & CreativeDone with custom-built or unnamed agent Oct 11, 2026
Turn AI newsletters into a daily podcast
An agent runs overnight, collects the author's AI newsletter subscriptions and produces a daily podcast to listen to on walks.
- What They Did
- In Claude Cowork on a Mac Mini, connected to Gmail, the author had the agent build a 5 AM scheduled task. It pulls newsletters under one label, skips about 18 low-value ones, reads the rest, and writes a synthesis. A local Qwen3-TTS model turns that into an MP3, which is pushed to GitHub Pages with an RSS feed. The author chose audio and the voice.
- How It Went
- It took about two hours in one session, and the first episode (8-10 minutes) was waiting in Apple Podcasts the next morning. Voice setup was the messy part: the macOS voice was unusable and edge-tts was unreliable.
- Worth Knowing
- Local Qwen3-TTS took about 30 minutes to install, with dependency conflicts and a stalled download, and peaked at 6.5GB memory on a 16GB Mac; the author used the Claude Max plan ($100/month).
Try it yourself with custom-built or unnamed agent
Every night at [time], collect the newsletters in my [Gmail label or folder, e.g. AI Newsletters] from the past 24 hours and pull out the key stories. Turn them into a [10-minute] podcast script with a conversational tone, then produce the audio and save it to [folder or podcast app] so it is ready for my morning walk. You are done when the episode file exists and you have sent me a short list of the stories covered.
Do this with a skill.
These SkillShop playbooks do this exact job.