tools for thought → tools for medium → medium for thinking → mediums for emergence
brainstorming
brainstorm what building a medium for thinking might look like
original ideas when working on obsidian ipfs are more along the lines of a tool rather than a medium (tools vs mediums).
I pictured, similar to are.na, that pieces of content would be blocks. Behind the scenes the blocks of content are actually ipfs blocks composed of any number of other blocks, each of which may be links to CIDs that could be fetched from wherever they were published.
Basically turning obsidian into a p2p protocols scrapbook of interconnected ideas. This would not necessarily be multiplayer, but would be collaborative.
I do think this is useful, but am wondering if maybe I should think about designing a very simple protocol , to support building convivial tools to foster usecases that are beyond the basic wiki/collaborative knowledge graph.
Perhaps along side the protocol, local first agents could become new types of “apps”, but more akin to a knowledge service. These agents could operate as doorless apps to your own data or personal vault. And could use UCANs to be given permission to draw new connections to parallel topics and even give permission to “curate” new ideas and topics into your vault of context.
Rather than the AI just writing new notes and links for you, it could perform a service of increasing your potential for drawing new connections by “depositing” new pieces of context into a partitioned area of your context vault that could then be read by other agents.
You could imagine an experience where for example you come across a blog, and an agent could display a graph or just list of other topics related in a more rich way than just keywords.
But with the use of agents, the outcome of examining context of the content could be much more dynamic and deep than current keyword or generalization of the topic,
Rather than summarization, curation, or roles of an algorithm to suggest new content based on the content,
Rather than the source of inspection and concept linking be based on the single static piece of content. The inspection could be changed to a much more dynamic intro/inter-spection of the context, links, and information related to the relationship between the publisher, the viewer, and the content.
Perhaps the agent could access the author of the blog’s context vault to see what areas of interest they are active in, make the connections and suggest some new way points that would not even be possible (tools vs mediums) by a human given the surface level perspective of viewing only a static page.
A single expression or output of a subject can’t encapsulate everything that surrounds it or influences it or accurately describe what it is, whether it’s ethos, broadness, multi-disciplinary, or hidden shortcomings.
This might be why we try to build these tools for thought. Our current tools of hypertext and links can only show snapshots of a a subject. We only see a single angle. It’s a single perspective of a typically static image, although with the web now does allow us to experience more dynamic views of things that are interactive and change over time (like general sentiment on issues in social media feeds).
This could also impact the way we engage with content as well. Currently we typically come across a blog or article or some piece of information and engage with it a single time. Of course we may bookmark it and refer to it later for a specific purpose. Our more long term repeat engagements over time are typically on apps and platforms, where things change over time by way of the content of the feed.
Apps are what we experience multiple streams content that we engage with once, but with a new model content can become the app or platform.
Rather than engaging one time, a piece of content or idea could more closely mimic our human connections and interactions.
Every time you revisit a piece of content (which in this hypothetical outcome) would maybe resemble more of a blurry concept than a static piece of content. “Content” would be closer to a group chat than say an image. Every time you revisit, the context of information surrounding it, including who/how many/sentiment of other people engaging with it would change. Your relationship to the author would change - maybe they have published new pieces of content that you have or haven’t engaged with which would have had butterfly effect types of impacts on any other pieces of content/concepts within the graph of concepts and content and agents.
privacy concerns
Once unfortunate outcome of this could be that maintaining separate identities would be more difficult to hide or keep separate to the general public than it currently is. Obviously having 2 accounts is prone to association by tech companies or the government, but generally the typical person online is not going to find the link if you don’t want them to. However in a world where agents are assembling context between points of data, it seems much more likely that the common, repeat associations of 2 identities across multiple content/contexts could become much more easily identified.
Sure you could have permissions and unassociated identities, but if your anon account is into the same type of topics and content or people as your public identity is, then it may be the same as seeing you at the same house parties over and over and every time you go into another room and come back out wearing a hat and different clothes or a disguise. it doesn’t become very hard to see the association between you and your alter ego.
what would a protocol look like to support something like this?
building of the starting point of content just being blocks and CIDs that link from other pieces of data, maybe each topic could be reflected in a “context node”.
The context could maybe be dynamic via somehting like roc or p2p resource computing. Where the context of a piece of data, the author, and the viewer would be passed to a node/agent that could assembles a new node of blocks containing all nodes that are N steps away and common across the 3.
for example, alice publishes a blog on the beach boys.
when bob views the content, a URI is requested of ipfs://BEACHBOYs@alice@bob
This could then view the context node of alice, of bob, and of beachboys. maybe producing the following links
- alice → music → instruments → traveling → african instruments
- beach boys → music → 60s → vietnam
- bob → friends → tim → foreign affairs →