Notes on Google NotebookLM's Audio Overview feature, which generates a podcast with two AI hosts discussing user-provided content and is surprisingly effective

the cadences, interleaving banter and narrative threads all of it is crraaazy [video] Deedy / @deedydas : NotebookLM may be Google's non-copycat product with organic traction in 15+ years with ~1M monthly users! Pixel / Photos / Gemini / Assistant / Waymo were all following other companies, Nest was acquisition, Glass / Stadia / Loon never really took off. Google should do more! [image] Andrej Karpathy / @karpathy : NotebookLM is quite powerful and worth playing with https://notebooklm.google/ It is a bit of a re-imagination of the UIUX of working with LLMs organized around a collection of sources you upload and then refer to with queries, seeing results alongside and with citations. But the Forums: Hacker News : NotebookLM's automatically generated podcasts are surprisingly effective

Simon Willison's Weblog 2024-10-01 Simon Willison

Context & Ripple Effects

NotebookLM had just added Audio Overview as a way to turn source documents into a two-host discussion, extending its source-centered LLM interface beyond written answers and citations. This early reaction matters because it identifies the audio format—not merely the underlying model—as the product element users find unusually compelling.

Subsequent coverage shows Google iterating on that format through customizable audio summaries and a business pilot, suggesting Audio Overview was becoming a broader NotebookLM surface rather than a one-off demo.

First-order effects

NotebookLM users can convert their own reading material into an audio explanation, offering a more passive way to review and make sense of uploaded sources.
Google gains an unusually legible product hook for NotebookLM: generated dialogue that demonstrates source-grounded AI output in a format users can quickly share and evaluate.

Second-order effects

AI research and note-taking products face pressure to make source-grounded results more consumable, not just more accurate; conversational audio becomes a competing interface for document synthesis.
The strong response raises the value of controls over tone, focus, and audience, a direction reflected in the later addition of customizable summaries.

Third-order effects

If source collections increasingly yield briefs, Q&A, and audio by default, the unit of AI work may shift from a prompt-and-response exchange to a reusable, multimodal workspace around supplied material.
The format could broaden access to dense material, but its usefulness will remain tied to users’ ability to inspect sources and distinguish fluent narration from reliable interpretation.

The trend: NotebookLM is one example of workflow-native AI moving from chat answers toward source-grounded, multimodal work products.

Discussion

@omooretweets Olivia Moore on x
The NotebookLM hosts realizing they are AI and spiraling out is a twist I did not see coming [video]
@thom_wolf Thomas Wolf on x
Self-care life hack: if you feel a bit down/tired, paste the url of your website/linkedin/bio in Google's NotebookLM to get 8 min of realistically sounding deep congratulations for your life and achievements from a duo of podcast experts 😂 [video]
@deedydas Deedy on x
Created this ~20min podcast from a 155 page report from Goldman Sachs on Generative AI in 2024 with Gen AI. Used NotebookLM for audio and Headliner for the video. [video]
@austinvhuang Austin Huang on x
NotebookLM is a reemergence of what made Google great. A few people got together to ship an idea that's cool + useful and then oops - PMF.
@giansegato @giansegato on x
uploaded my last essay to notebookLM. it's... uncanny. some thoughts: - first time i'm coming across AI content that feels genuinely human - it doesn't simply summarize the content, it adds a meta-commentary on top, exactly as a podcast host would do; this is 1. new behavior, 2. …
@swyx @swyx on x
@karpathy very googley to accidentally drop SOTA TTS as a side project of something titled “notebook”
@lateinteraction Omar Khattab on x
The way NotebookLM generates podcasts under the hood is a modular system quite similar to STORM. [image]
@giulio_matt Giulio Matteucci on x
NotebookLM is really fun! I've integrated the podcasts it generated for my papers into my personal website ( https://www.giuliomatteucci.net/ ...). While there are some minor inaccuracies, it's impressive how well it conveys the key findings in an engaging and accessible way!
@swyx @swyx on x
just realized NotebookLM is @GoogleDeepMind's ChatGPT moment - “low key research preview"/"experimental" - not monetized - GPUs/TPUs immediately on fire - SOTA proprietary new model buried in there with upgrade that weren't previously announced - new AI UX that cleverly embeds [i…
@panzer @panzer on x
Gave the NotebookLM AI hosts some info about @tiptopxyz and they had a really interesting chat about it for 8 minutes and change. Absolutely mind blowing veracity here — the cadences, interleaving banter and narrative threads all of it is crraaazy [video]
@deedydas Deedy on x
NotebookLM may be Google's non-copycat product with organic traction in 15+ years with ~1M monthly users! Pixel / Photos / Gemini / Assistant / Waymo were all following other companies, Nest was acquisition, Glass / Stadia / Loon never really took off. Google should do more! [ima…
@karpathy Andrej Karpathy on x
NotebookLM is quite powerful and worth playing with https://notebooklm.google/ It is a bit of a re-imagination of the UIUX of working with LLMs organized around a collection of sources you upload and then refer to with queries, seeing results alongside and with citations. But the

Chronicles