Gemini researches cultural history for a podcast
Brenner reports using Gemini Deep Research for cultural-history reports that became public narrated episodes, with a separate Python pipeline preparing and assembling the audio.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Research [specific cultural-history question]. Develop an investigation plan, follow primary and reliable sources, and produce a report with citations. Flag uncertain links between people or events so I can review it before adapting it for narration.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Ben Brenner
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Brenner reports using Gemini Deep Research for cultural-history reports that became public narrated episodes, with a separate Python pipeline preparing and assembling the audio.
- Human involvement
- Brenner chooses questions, builds the production pipeline and replaces an unreliable all-in-one approach. The source does not establish unattended agent publication or independently checked historical accuracy.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Give Gemini Deep Research a specific cultural-history question.
- Inspect the agent’s plan and resulting sourced report.
- Export the report through Google Docs into the Markdown production pipeline.
Ben Brenner asks Gemini Deep Research questions whose answers lead into unexpected cultural history. A question about filmmaker Mike White develops into research on Wesleyan alumni's influence on film and television. A question about kolaches reaches Czech immigration and Cambodian-American donut businesses. The agent plans the investigation, finds material and synthesises a report.
Brenner publishes each episode in two forms: the original research article and an audio version. The downstream production system is deliberately separate from the research agent. He exports the report through Google Docs, ingests Markdown in a Docker container and uses Python to divide it into sections. Gemini Flash rewrites tables and formatting for spoken delivery; pronunciation substitutions precede paragraph-by-paragraph text-to-speech and audio assembly.
An attempted single language-model chain was too inconsistent, so Brenner replaced it with small tools. This is not evidence that Gemini autonomously published a podcast from one question. The completed public article-and-audio pairs demonstrate the combined workflow, while historical accuracy still depends on the report's sources and editorial judgment. The useful distinction is to preserve the readable research report while adapting a separate narration script, particularly when tables or pronunciation make a literal reading awkward.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Blog · Ben Brenner
Frequently asked questions
What makes this more than text-to-speech?
The research agent investigates a question first, and the production pipeline then adapts tables, section breaks and pronunciation for listening.
Why retain the original article?
Brenner publishes the research report beside the audio, allowing readers to inspect the written material instead of receiving only narration.
Was one agent doing the entire process?
No. The research agent is followed by a separate Python pipeline with rewriting, text-to-speech, assembly and hosting tools.
More details
Remaining steps
- Adapt tables and pronunciation for speech, then assemble paragraph audio.
- Publish the readable report alongside its narrated version.
About this account
Gemini Deep Research plans and performs the investigation; separate Python and model-API steps adapt and assemble the report into podcast audio.
Firsthand account by Ben Brenner; original publication date verified against Primary blog date Aug 8, 2026. Outcomes are the author’s reports, not independently reproduced. Suggested prompt written by Agents In Use; not the author’s original prompt.
Credit: Ben Brenner
Source published
Added to Agents In Use
Source checked
Tools and topics
Gemini Deep Research · Google Docs · Python · Docker · Gemini Flash · Gemini TTS · Cloudflare Workers
Cultural-history research · Report narration · Audio adaptation
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.