Creating a synthetic podcast on United States v. Skrmetti

The Supreme Court decided United States v. Skrmetti this morning. The 6-3 decision upholds Tennessee's ban on gender-transition hormones for minors, finding no sex-based classification and applying only rational-basis review. In light of the decision, I created a synthetic podcast discussing the case. You can listen to it here. Or, if you don't have 22 minutes of listening pleasure in mind, here's the 6-minute version.

As to how I did this within 30 minutes of the opinion distribution, this was the process. As discussed in an earlier blog entry, I had AI (the Legal Case Briefer CustomGPT on ChatGPT) read the opinion and brief it. Since I expect some inquiries from local press, I had AI draft a list of ten questions one might anticipate from Texas journalists. I asked for a one sentence answer to each question (for TV) along with a paragraph length response (for press or internet journalists). I then asked for a "Linda Greenhouse or Nina Totenberg style article on the Skrmetti decision. 1000 words." That "article" advised me of something I had not previously considered. So I followed up with the AI.

You wrote: "That framework may prove significant well beyond transgender health care. In recent months conservative lawmakers in Texas, Florida and Missouri have introduced bills restricting access to fertility services, cosmetic surgery for minors and even certain forms of contraception. If those measures are framed as age- and diagnosis-based rather than gender-based, Skrmetti supplies a ready blueprint for withstanding equal-protection attack." Tell me more about that.

The AI dutifully produced an answer. Because ChatGPT does not make it easy for other AIs to access full chat transcripts, I then copied and pasted each of its responses into a Google Doc. You can see the 15 pages of results here.

And from here, the process was easy. I just launched NotebookLM, created a new notebook that it dubbed "Skrmetti: Remapping Equal Protection and State Medical Policy" and asked it to create a podcast. To make it interesting, I asked it to give one host a moderately liberal perspective and the other a moderately conservative perspective. And that was it. A 22-minute synthetic podcast emerged. I know that's too long for many busy people. Since NotebookLM does not currently support multiple audio overviews within a single notebook (Google, please!), I then created a duplicate notebook and asked for the short podcast version. A typical six-minute radio segment was the result.

There is, of course, much more that AI can do. That's what this blog is devoted to. It can create a presentation based on the decision. It can create a study guide. You can ask it to rebrief the case from a particular political perspective. Indeed, here's a feminist perspective brief of the case. But for now, listen to the podcast and think about what the court has done. And think about how you can easily create a podcast of your own any time a recent court decision interests you.