Peter Yang’s Three-Step System for Automating Podcast Production

Peter Yang explains how he uses a chain of AI skills to produce his weekly podcast. He estimates the system saves at least five hours a week, while acknowledging that he still spends time reviewing and editing. The tutorial is sponsored by Riverside, which he uses in the demonstrated workflow.

First, he maps the manual process in detail: guest research, an interview guide, recording, an editor handoff, review, publication, show notes, newsletter, social posts and clips. His list has 15 steps grouped into preparation, editing and post-production. He asks AI which steps it can assist with.

Second, he creates reusable skills for those stages. A preparation skill researches and structures the interview; an editing skill prepares instructions for the video editor; a production skill coordinates specialist skills for titles, show notes, the newsletter, social posts and clips.

For an interview with ChatGPT Finances product lead Ethan Bloch, Yang asks the preparation skill to research the guest, scan related videos and propose an angle, questions and packaging. He rejects some suggested topics, asks for more practical use cases and revises the interview guide before sharing it.

After recording in Riverside, Yang uses a Riverside MCP connection and his editing skill to prepare a Linear ticket. The assistant proposes intro-reel selections, cuts and clips from the transcript. Yang combines alternatives, adjusts clip boundaries and reviews the instructions for the human editing team.

In post-production, a newsletter skill drafts the episode write-up. The initial version misses or distorts some prompts, so Yang supplies the six prompts he wants and requests a revision. His example shows that the first draft still requires fact and content review.

A thumbnail-and-title skill analyzes the transcript and researches comparable videos, then proposes several combinations. Yang chooses and refines options for testing. He says a repeatable thumbnail template and a Figma connection can also reduce manual work.

A show-notes skill drafts descriptions with timestamps, sponsor information and links. Yang asks for more specific chapter labels and uses browser automation to place the approved description and configure the YouTube upload. He still checks the result.

To improve the newsletter skill, he snapshots the draft before making manual edits. He then asks the assistant to compare the before and after versions and propose numbered skill changes. Yang reviews the proposed changes and gives feedback before asking for updates.

A social skill uses examples of his previous writing to draft teasers and posts. Yang edits the opening and chooses a suitable screenshot before scheduling a post. The examples and his review are part of the workflow, not an assurance that the first output will match his style.

For clips, the assistant suggests moments from the interview. Yang selects one, uses Riverside to create and review the clip, checks the accompanying copy and schedules it through Typefully. The workflow combines connected tools with decisions by the host and editor.

To adapt the method, Yang recommends choosing one or two time-consuming workflows, listing each manual step, creating reusable instructions for recurring stages and connecting them in sequence. A skill should explain when to use it and how to perform its steps.

He closes by noting that even this system still takes substantial time and that he may need a person to help operate it. His claimed time savings are specific to his own process and do not imply that every workflow can be automated to the same degree.