A beginner's guide to building a weekly digest

I listen to a lot of podcasts. And I mean a lot. Most of them are for me. The personal interest stuff. The industry insights, the AI deep dives, the tooling conversations are the ones I keep meaning to get to, but they pile up like the one chair in the house that somehow has a bunch of clothes that for whatever reason can't seem to find their way to the closet.

Reading is random-access. I can skim, jump, search, go back. I can read one paragraph out of an hour-long conversation and decide in six seconds whether the other fifty-nine minutes deserve my attention. Listening is sequential. It assumes I've agreed to hand over the full runtime, and lately I haven't. But I still want the signal.

So I built a weekly digest that lands in a private Slack channel every Sunday and tells me what AI, product, and tooling podcasts talked about this week. Where they converged. The specific tools and companies people mentioned. The verbatim quotes worth stealing. The "I should probably build this" and "I should probably teach this" signals I would have picked up if I'd actually listened. Skimmable at the top, drill-down in the threaded replies if something catches my eye.

This used to be a committed amount of time to build. Pulling feeds, parsing subtitles, structuring LLM output, writing a Slack formatter, handling rate limits, scheduling the whole thing. Now it takes an hour.

We're drowning in information. The problem was never access, it's attention. It's why I use AI to find the signal. The surprise isn't that this works. The surprise is that building something this specific, to my taste, just for me, takes such a short amount of my time.

One of my kiddos saw the podcast digest I get every Sunday morning and asked if I could do one for him for philosophy.

He's studying philosophy in college. He sent me a list of sites and his interest. He wanted "tell me three things to read this week, and a paragraph on why." And for the reference works (the ones that don't publish), he wanted something more like a study plan. Pick a topic. Tell me where to start digging.

The thing I built for him has two sections and is delivered to him weekly in an email.

The top half is "what's new," pulled from the actual feeds. Three picks, a paragraph each on why this one and not the others. This is the part that looks like my podcast digest.

The bottom half is "this week's topic." There's a list of 15 topics and the agent rotates through them and generates a plain-spoken overview.

I am not an engineer. Using AI this took me about an hour. It's completely free. Free tier Gemini, free tier GitHub Actions, $0 a month.

For most of my life, "I made you a thing" meant a mixtape, or a playlist, or a binder of recipes printed at Kinko's. The bespoke gesture was always shaped by what you could realistically build in a weekend with the tools you had.

The shape of that gesture has changed. I made my kid a weekly personalized philosophy digest, because he asked. That is a different category of thing than a playlist. It runs. It will keep running. It is, in some small way, an ongoing piece of attention I get to send him every Sunday morning without being weird about it.

If you want to try to build something personalized, here's a repo to get you started: https://github.com/bonus414/weekly-digest-template

Happy building!