Stock, Flow, and Chat
4 min read

Stock, Flow, and Chat

Stock, Flow, and Chat
Big. Powerpoint. Energy.

Authors Note: As I slowly withdraw from Twitter and return to blogging, I'm going to share posts that I would've used Twitter to compose and publish. Consider this an attempt to share a rough idea for feedback. This is raw and I'm guaranteed to be wrong but there's potential to figure things out by sharing and talking. Appreciate anyone taking the time to read! -Noah

Nearly 12 years ago, Robin Sloan (Personal Website) applied the economics concept of "stock and flow" as a metaphor to explain early 21st century media. He wrote (emphasis mine):

Flow is the feed. It’s the posts and the tweets. It’s the stream of daily and sub-daily updates that reminds people you exist.
Stock is the durable stuff. It’s the content you produce that’s as interesting in two months (or two years) as it is today. It’s what people discover via search. It’s what spreads slowly but surely, building fans over time.
Flow is ascendant these days, for obvious reasons—but I think we neglect stock at our peril. I mean that both in terms of the health of an audience and, like, the health of a soul. Flow is a treadmill, and you can’t spend all of your time running on the treadmill. Well, you can. But then one day you’ll get off and look around and go: oh man. I’ve got nothing here.

As we've evolved from blogs to tweets and back to blogs (aka newsletters), I think Robin's central conceit still holds remarkably accurate. Yet, are things about to evolve?

With AI – and in particular ChatGPT – are we about to see a flood of new objects adding further noise to the flow and/or will we see the creation of new durable artifacts resisting the feed's current to join the stock?

Last week, Dan Shipper (🐦 Twitter) documented his process building a chatbot leveraging the archives of the Huberman Lab podcast archives. Dan's bot was able to leverage thousands of hours of audio conversations to create useful tool that not only summarized, but attempted to (somewhat successfully) synthesize complex concepts into brief and cogent prose.

I Built an AI Chatbot Based On My Favorite Podcast
Here’s how I built it and what I learned about the future

While reading Dan's post, I kept returning to Robin's stock and flow metaphor. What resonates most with me about Dan's enthusiasm about AI powered Chatbots is that we're seeing the potential for new forms of stock.

For the last decade and change, most new types of content – Snap Stories, TikToks, etc – usually wind up adding more and more stuff to the flow. If we let ourselves, we could easily drown in our never ending feeds. To be clear, I think the growth of AI powered synthetic media will function to create an abundance of additional flow (especially with images and videos).

What I didn't expect was to see the potential of new forms of stock to also emerge just as rapidly as these new forms of flow. Go back to Dan's Chatbot built on the Huberman Lab podcast archives. The Huberman Labs is one of those sprawling conversation podcasts that delves into complex topics resistant to easy summarization.  It's a podcast built on core ideas that evolve over time.

As a media format, it's the type of podcast that exists as both flow and stock. The cadence of fresh episodes (aka new conversations) keeps nourishing the feeds while simultaneously the core ideas being discussed over time are a form of stock. A challenge inherent to this audio format (podcasts) is that the stock is difficult to retrieve without repeat listens. Even if you had access to pristine transcriptions, it's hard to follow what's being discussed without starting from the beginning (or an earlier conversation).

AI powered chatbots offer a potential antidote. What if you could offer a reader a complex idea built off of synthesizing thousands of hours of podcast transcripts? What if you could encourage a reader to keep asking questions by drilling deeper and pursuing unexpected rabbit holes? Isn't this stock? Isn't this like going on a Wikipedia tangent only to find yourself reading tabs from the archives of the New Yorker or Atlantic?

Now what if this bot wasn't limited to any one podcast. What if it was mining the entire archives of every American newspaper from the last 150 years (and was updated in real-time as new reporting is published)? I'm so curious about this new interaction format as both a potentially new form of stock (the chatbot conversation) as well as an entry point for people to find existing stock (discovery of URLs via these chatbot conversations).

Time will tell, but I'm eager for some fresh distribution.