Dale Spoonemore
← All posts

Writing · Dale Spoonemore

I’ve been turning what I need to learn into podcasts

November 13, 2026

Mint-colored waveform from the example podcast against a dark teal background.

Listen to the full podcast (1:51)

The full edited episode about this process. AI-generated voices from NotebookLM, edited for accuracy. No company documents were used.

When I joined Untangle, I had several unfamiliar technologies and a new industry to learn. There was a lot to get up to speed on, and one of the most useful things I started doing was having Devin help me turn that context into podcasts I could listen to.

I’d have Devin put together context files and scripts or prompts based on the actual project, then take that material into NotebookLM and have it generate a podcast. I started doing this during onboarding, both to learn the project and to get daily updates on what Ryan was actively developing. It gave me a way to work through a lot of unfamiliar information in a format I really enjoy.

The important part of this process is where the context comes from. Devin can go to the source of truth and put together the material about what I’m trying to understand. That gives NotebookLM something specific to work from. I want the explanation grounded in the project itself, with the context coming from the agent working with it.

That’s been useful well beyond onboarding. Now I use the same process when I’m trying to understand a particular process or something going on in the app. I can focus the material on the thing I need to learn instead of starting over with an introduction to the whole project.

I also like having different versions of the same topic. I’ll ask for a short overview, somewhere around five to ten minutes, a version that’s roughly twenty minutes, and a deeper dive that might be forty-five minutes to an hour. Those are the lengths I’m aiming for, depending on how much detail I want.

I start with the short one. Sometimes that’s enough to get my bearings. If I need more context, I can move to the longer version and spend more time with it. I don’t always know how much detail I need before I start listening, so being able to go deeper on the same topic has been really useful.

At first, I was moving the material between Devin and NotebookLM manually. Devin would prepare it, and I’d take it over and get the podcast generated. Since then, I’ve automated that handoff with Codex in the loop. Devin communicates directly with Codex, and Codex drives the browser to generate the podcast. The basic process is the same, but I’ve been able to automate the part where I was moving things between the tools.

There’s a pretty simple idea here that I think other people could try: start with a specific thing you want to understand, have the agent with access to the relevant sources put together the context, and use that material to create something you can listen to. Start with an overview and ask for more depth where you need it. The quality of that starting material matters because it’s what the explanation is built from.

NotebookLM’s podcasts sound remarkably real to me, and they’ve become my favorite way to learn. This has fundamentally changed how I take in information and helped me get up to speed quickly. I’m still using it as new questions come up, and I really like being able to turn the thing I’m trying to understand into something I can listen to at whatever depth I need.