The Arcology Garden

Fediverse Second Brain


what if instead of being weirded out that you have to pick a server when signing up to the Fediverse you and your friends could be shit like or whatever for nominally 0$ a month, or maybe like 5$ a month after you start backing up the homeserver to backblaze

what if knowledge work was accomplished as a conversation with a set of actors, some programmatic, some probabilistic, some human?

i got stoned and wrote some long threads where i think about a microblog interface for the arcology project/knowledge management/notetaking and this sort of "iterative thinking in public" pattern that i have between Fediverse and org-roam. A simple realization is that right now my thinking flows from post->org->post but my tooling only goes org->post. here's how you could over-engineer a solution to that and build a weird public second brain which allowed 3rd party thinkers. here's how you could build a Concept Operating System.

consider a tailscale+funnel hosted microblog appliance where the public view is available over the internet on funnel and the posting interface and any higher level tools are authenticated by the tailnet.

what if instead of having a mastadan profile with a linear scrollback of your posts, the profile view was a personal web site you could customize, and import/embed posts and threads in to it from your client

what appears to just be an activitypub server based on bonfire or so with a "related posts coalescing/search" feature could become a compelling collaboration tool.

INPROGRESS a second brain with multiple interfaces

microblogging starts to get weird when you add the public site + ability to "self-canonize" or re-summarize certain trains of thought, and then create wiki-like sites where you can follow links and backlinks between topics.

so the public view of your profile is a web site not a profile page, an enriched view of the fediverse posts, pages assembled by thread or maybe CW/summary, and using the vector db or full text engine to group similar threads together and help a user generate useful indices and distill knowledge from long term thinking or stream of consciouness shitposting on a subject in to a bunch of org-mode headings in a single page, which can then be iteratively turned in to a more polished post, like what I am doing here; instead of using Feediverse to publish the whole thing, documents like these could be parsed and deconstructed in to threads which can be posted to an outbox on the activitypub thing.

so you have a activitypub for stream-of-consciousness, and a tool that lets you distill those ideas and re-present them.

and that thing can be programmable and generate literate programming documents

probabilistic and programmatic agents

and if that thing is programmable well, it can have agents.

if the activitypub server is not just an "off the shelf" thing like mastodon or akkoma or even gotosocial, but something built in an actor system like elixir/bonfire libries/etc you could have self-hosted programmable agents.

and so you have the ability to message automated agent actors in the system (and develop new ones).

query the db by tagging an actor; tag a CI runner in a post where you upload a .patch to your coworker after a long discussion on a potential API design; tag the same bot later in thread with some code or an full actor process (elixir genserver, liveview module, etc) and it tangles the code out to a filesystem and replies with a report on the build output and offers to run a new actor in the active VM

This could all be an un-federated and only available to machines on your tailnet. but maybe you get a sort of multiplying factor if you have certain ideas slowly open to consideration of your wider peer group.

iteratively thinking in public

so if it was federated, even if it was only federated with a few friends over tailnet, you could do a sort of weird "thinking in public" thing where you start by developing a thought in a DM to yourself, and then maybe loop in your collaborators, maybe they are on their own instance, maybe the posts are local-instance only to protect trade secrets (lol). you two get excited by an idea and summarize it in an boostable post that goes viral. you get a lot of feedback and implement the idea, you slurp all the posts from you and your collaborators (which now could include the public) into an org-mode doc and write a final version of it as a literate programming doc which can be hosted on the public site

literate programming of this thing is important, to me, especially getting it to be self-hosted or at least can self-host extensions; like after installing it in a VM or on a host, you should end up with a link to a site that has instructions on setting up the client, etc, pre-populated in the @admin account and published to the public view w/ the ability to extend that page and the functionality itself documented by posting notes containing code or configuration

silly moonshot/Hypergoal product idea

eventually you build a product that is a VM image or an SSD that you plug in to an 100$ tower and hook a monitor up to and plug in. it boots up running