For some time now, I’ve wanted to create a system that would allow anyone to credit anyone for anything. I had the intial idea years ago, registered the https://cred.at domain, then just sat on the idea.
The intital idea was for a dumb simple, single line of text where you would do something like:
lux@cred@mom "best apple pie ever"In this case, ‘lux’ would have given credit to ‘mom’ for making a great apple pie. At http://cred.at/lux/given, lux would see a list of credits he gave, and at http://cred.at/mom/credits/roll, mom would be happy to have credit for the great pie, and all the other credits.
I shelved this idea because it felt to broad, too stupid, too many issues with implementation, etc, etc, but I never was completely able to drop the idea.
Later I thought about using cred.at as a sort of “belief network”, that would use natural language processing to determine attributions, so things like user@cred@historical_figure might be possible. For instance, the nlp processor would produce things like lux@cred@columbus “discovered America” while another user someoneelse@cred@vikings “discovered America”. This would be a way to track beliefs and attributions though time, to understand why people believed the things they do, and where there is divergence. I think this is still an interesting idea that someone should definitely pursue and be implemented somehow, but again, it felt to broad.
Recently, especially in light of data needs for the AI/ML revolution, I’m seeing a need to have a way to attribute across systems.
For instance, you might want to attribute a research paper in arxiv, but implicit in that attribution, you are attributing the work of several authors, you are also attributing the work in their github repo, some related tweets on x (formerly twitter), some bluesky posts, etc etc.
There’s also need to attibute sources in ai model training data as well, and a need to attibute artists for their work, etc etc. Also, for the purpose machine understanding, it would be great to have a system where the LLM could easily track down the root of an idea, tracing, though the whole historical attribution chain in one fell swoop.
All the current systems for doing this are locked into one platform or another: there’s issues of data control, verification, deleted information, etc.
This is why I think it’s time for a system that is universal, open, distributed, optional, verifiable, platform independent and dead simple to use. I hope to make that system a reality here.
I’ve started a github repo to start building out the protocol syntax and intital backend (don’t get too excited: still just a README doc, but go star it anyway).
I will be building the first implementation at https://cred.at, that will allow users to link their identies on various platforms and have a universal “inbox” for recieveing cred@, but my goal here is to make a system where anyone can implement their own cred@ system if they like, and it will be interoperable with any other.
If you are a researcher, developer, or just someone who wants to give a cred@ to a friend on twitter, and are interested in being an early adopter, definitely reach out and let me know. It’s totally minimal effort, you’ll just get an account on cred.at, a place to host a blog, and the ability to give and recieve cred@.
If you understand the implications and value of a system like this, early stage investment is also welcome …in either case, feel free to reach out to me on github, twitter or bluesky.
Thanks for reading! I’m also going to post some amazing AI/ML ideas on this blog as well, so visit again soon. Here’s the general outline:
cred@ is an open, distributed attribution system designed to create verifiable chains of intellectual credit across platforms and mediums. It provides: - Universal syntax for attribution - Verifiable credit chains - Automated tracking - Distributed verification - Platform independence
The system allows both passive tracking of attribution and active user engagement, while remaining platform-agnostic and open source.
Core required elements:
cred@system:resource:unique_idFull syntax with optionals:
[creditor]@cred@system:resource:id[:user][:temporal]system: The platform/source (arxiv, github, etc.)resource: Type of thing being credited (paper, repo, tweet, etc.)id: Unique identifier within that systemuser: Optional specific attribution (defaults to all contributors)temporal: Optional version/time reference (defaults to latest)# Basic paper citation
cred@arxiv:paper:2311.12345
# Specific version/author
cred@arxiv:paper:2311.12345:hinton:v2
# DOI reference
cred@doi:paper:10.1038/nature12345# Repository
cred@github:repo:user/project
# Specific file
cred@github:file:user/repo/path/file.py
# Specific commit
cred@github:repo:project:user:commit-a1b2c3# Tweet reference
cred@x:tweet:123456
# Blog post
cred@medium:post:user/post-id
# Video content
cred@youtube:video:video-id:timestamp# Digital art
cred@opensea:artwork:token-id
# Music
cred@spotify:track:track-id
# Photography
cred@flickr:photo:photo-id