Pubsub could be reframed as a principle for innovation in consumer tech.
Pubsub (short for publish/subscribe) is a pattern where publishers shouldn’t have to know who they’re publishing for, and subscribers shouldn’t have to know who they’re subscribing to. Users just have to care about a topic.
Examples from social media, crypto, and AI:
Marketplaces
In a marketplace, a buyer doesn’t have to care who the seller is and vice versa. Users just look up the item they want and click buy.
Peer to peer marketplaces wouldn’t scale at all without pubsub, and iterating on this principle allows them to scale further (see Web3 and Generative AI below).
We take marketplaces like these for granted today, but they were a huge innovation for the internet. I highly recommend reading How the Internet Happened by Brian McCullough.
Social media
Content creators in social media don’t have to manually distribute content to each of their consumers. They can publish content widely, without having to whitelist each user allowed to find or consume it
Consumers also benefit because they don’t have to request access to each piece of content. They can just find photos of their friends or of a public figure they care about in real time. (These are the “topics” for pubsub.)
Search engines
Search is beautiful because it creates a pubsub system out of disjoint content:
- When a publisher creates content, search engines index it by tagging it with topics (text, alt tags, meta tags, and other kinds of SEO tricks). Publishers don’t have to explicitly guess the topics that subscribers will look up, now and forever into the future.
- Consumers type in the topics they care about, and search engines rank content under the closest topics in their index.
This innovation enables content to decentralize. Search engines like Google create centralized indices for external content and users visit a centralized index before they arrive at the content.
Web3
In web2, we have middleman companies acting as pubsub topic repos, and consumers find topics they care about by first looking up the repo and then the topic.
Web3 allows us to just focus on topics, and have companies compete to provide the best experience.
An easy example is NFTs. Multiple marketplaces are finally able to share the same inventory, because the source of truth (of who owns what and how to change that) is independent from them all.
Generative AI
On the current web, the topics that subscribers care about are actually distinct from those that publishers define. Search engines make a best effort at matching the two (see #2 above).
We can now fix this using semantic search and by generating personalized content. Users can subscribe to the topics they actually care about (not the ones the search engine might be good at) without having to know which topics publishers care about.
AI then creates the most appropriate content, regardless of the topic (language, website, image, etc) used by publishers.