For many of my posts – most, these days – I use NotebookLM to generate an audio summary in the form of a podcast: a dialogue between two virtual hosts. Some listeners have complained, but I stand by the practice.

First, some people prefer to listen rather than read. They might be driving, cleaning, or simply allergic to text. I see no moral failing in that.

Second, the virtual hosts do more than recite; they interpret. They summarise, add perspective, and occasionally introduce sources or explanations I hadn’t included. The quality varies – some episodes hit the mark, others wander into creative misreading – but that’s no different from human discourse. When they err, I consider whether my prose invited confusion. If so, the fault may be mine.

And yes, if you dislike AI-generated audio, you’re free to skip it. I can’t provide that perspective myself; I’ve already written the piece. I could, I suppose, rework my essays to address their objections and then pretend the machines weren’t involved, but where’s the honesty in that?

Finally, some people only encounter my work through these podcasts. They rarely or never visit the blog, yet the ideas reach them all the same. The blog and its neglected companion YouTube channel now have the widest reach. I’d like to produce more video content, but editing devours time. For now, NotebookLM carries part of that burden, but I’ll be taking up some of the slack soon. Probably.

---

EDIT: Funnily enough, in the audio summary, NotebookLM is suspiciously unaware that it is evaluating itself – though it does seem to push some self-promotional angles.