Web3 Ideas: Interactions compound towards Computability

Send your thoughts via twitter or mail

stage: early draft

this post gets somewhat technical in places

The web has caught up with everything operating systems can do.  The web runs most of the applications in the world. It takes care of account management, rendering, GPU access, voice calls, device abstraction (webcam, microphone), USB access, VR, and more.

Browsers have become a stable, backward-compatible platform. Most apps are becoming web apps. The browser is involved in most of what we do.

So as an idea, what if....

...the browser captures (most of) your interaction data so you can leverage it for smart agents or expose it to services through a finely grained, permission-ed programmatic interface.

Capture what interactions?
Everything you do on the web: books you buy, how long you read some article, the paths that led you to a piece of content, what products you pay for, the tweets you read (and skip), what web apps you use and for how long, when you started following an account, how much time you spend on news sites, how many emails you send a day.... that’s all interactions that you currently don’t own.

The browser can harvest data produced by any online experience (sites, web apps, messages, zoom calls etc.) as long as there’s a sort of event adapter1 for it. At the moment, no interactions of that sort are captured for us to work with2.

Why would I want this?
You could leverage every action you make on the web into a possibly compounding asset. The decisions you make form your interests, tastes and preferences. Social media companies make billions owning a small slice of that data.

You can use those assets to compile content recommendations, get notifications when you watched the same video as a friend, find a group’s overlapping tastes with your reading history (find intersections and unions), run a poor-man’s emulation of you3, pay you for a slice of data for a population study, autogenerate a newsletter4, spawn a web crawler that finds job posting that fit you, overlay smart content-filters over social media feeds, receive targeted marketing against payment, generate draft email responses, rent out slices of your store to 3rd party services or a government, receive context-aware annotation gigs, solve cold-starter problems with whatever the next Netflix, meetup.com, Twitter is, compile personal analytics on reading habits...and a million other ideas that a market can get creative about.

Many jobs that are currently done by $M or $B apps might consolidate into a few graph queries with a standard UI component slapped on data that comes back5. Whenever an interface captures a glimpse of us, we should own it.

This sounds weird. Where is this going?
Let’s start from the future and work our way back. Software really is eating the world and the world makes itself more edible to software.

With that, large chunks of reality move into databases. We are becoming increasingly computational wherein humans are expected to sponsor the same affordances as computers. Unfortunately, Human-computer interaction (HCI) innovation has slowed down after Xerox Park (GUIs, mouse, touch pads) and so we are forced to become like computers from the 70s instead of computers becoming like humans from 2021.

That forcing might lead to people having API endpoints that agents with permissions (governments, 3rd party services, chat group members, coworkers, friends, institutions) can query against...enabling actual peer-2-peer messaging.

This removes the need for presence and abstracts a slew of lower-level cognition, ie. aggregating and filtering over simple entities5.5. Suddenly, group coordination like a multi-book club meeting with strangers in a new city is a hundred! times less effort.

To some extent we already sponsor real-time APIs through devices. Ride sharing apps query your location for example. Yet nobody can ask questions across devices, apps, histories  (ie. multi-hop queries) and more importantly: groups of people. That’s because every piece of data is owned by someone who isn’t you and scattered through a bunch of private, venture-funded databases.

Alright, that was really weird...
Yes, right now it all this seems frivolous, but capturing and hosting your computer-readable data is a key part of becoming a distributed humanoid6 species.

OK bye.


[ TODO: Idea 2; Every entity type (Person, Stock, Tweet, Video, Article, Company, Book and so on) is a category in a marketplace of serverless functions and services. ]


[1] Event Adapters: In web apps that’s attaching event listeners to elements in the rendered document - using the DOM. These are hooks into the browser (possibly as extensions) which will map arbitrary user actions to meaningful datums. For example [clicking that button on this domain (ie. youtube)] maps to [upvoted a video]. That datum then goes into3  the user’s private, local data store.

[2] I had to write a script just to export browser history and a few more scripts to answer “What articles did I spend time to read this month?”

[3] depending how many data points... in general the more behavior is digital, the easier it is to digitally clone someone’s behavior

[4] Initially just simple list-like things like “best of Twitter”, “month’s favorite articles”, “most played songs”, or my “recommended” collection

[5] Like in reactive programming, services would subscribe to a subset of your state

[5.5] Like Person, Stock, Tweet, Video, Account, Article, Company, Book and so on

[6] not my choice, but that’s what tends to happen when apes start playing with transistors