I think your model certainly makes sense for some use cases. It’s totally valid to create a local graph based on some local understanding and work off of that graph for some use cases.
But the mental model and value of having a global graph using entities with stable IDs is different.
Allowing humanity to grow its knowledge over time. The only way to truly grow knowledge is to have something stable that grows and accretes. You can’t do this if the underlying targets of description are constantly changing
Enabling communication at a global level. Think of things like language. How hard would it be to communicate if words meant different things in every conversation? You would have to constantly maintain a view of what each word meant in each sentence in order to understand anyone. This is a Tower of Babel situation and communication would break down.
This may provide some intuition for the value of having a global graph that is semantic so that we as a species can accrete knowledge over time and communicate across applications and communities.
Hey Yaniv and team, thanks for sharing and building this!
In the spec you say Solving the remaining composability challenges will enable an interoperable web3, I cant help but think this is much bigger than web3, it could enable interoperable knowledge across … well everything really. Super hyped about GRC-20.
I have a few questions/thoughts after reading the spec:
The specification seems quite simple and elegant and the problem of interoperability has been a theme in the web3 community for a while now. Do you know of any other initiatives/projects that attempted (and maybe failed?) to create something similar?
The document mentions Knowledge is published onto The Graph using onchain transactions. Maybe I’m misunderstanding this, but are there plans to have a cannonical blockchain where public knowledge is stored? What does that look, a set of smart contracts probably? And I assume that will be seeded with a bunch of “top level” entities like those described in the spec that already have an assigned ID, is that correct?
Any plans on building documentation or maybe tools for developers that want to use/integrate with the standard? For example, let’s say I’m looking to build a dapp that uses GRC-20 based knowledge graphs, where do I start?
Related to previous question, it would be nice to have a sandbox environment with a pre-populated (small) knowledge graph to play with, I know as I was reading I wanted something like that to exist to visualize and explore the ideas in the spec. A simple local first website would be nice.
How much of this is implemented on Geo?
Thanks again, excited for the future of The Graph!
BTW, I was already sold on the idea then I came across this post by @0xThierryx.com, brilliant stuff.
Driving AI agents’ behavior and governance with easily swappable knowledge graphs is a great way to empower people and take back some control, which is what web3 is about no? I wonder how far this can go, maybe there’s a future where knowledge graphs are incorporated into AI training as a way to reduce trainer bias or at least steer it toward individual preferences. Very interesting!
Agree! GRC-20 could be used anywhere. But I also think everything will migrate to web3… so that’s why I sometimes say that GRC-20 enables web3.
There are some projects trying to do knowledge graphs for web3. Intuition seems to be one on Ethereum. OriginTrail is one on Polkadot. Golden tried to do something but I believe shut down. I mentioned in Knowledge graphs are web3, the 3 main reasons I think attempts to do something like this have failed: 1) RDF is a complex and imperfect standard 2) Challenges with governance for shared data and schemas 3) Not having the right incentives and economics. I would add a 4th which is not having an intuitive interface for browsing and contributing to the knowledge graph and tools for 3rd party developers.
Yes, there are plans for this but we haven’t described them yet. We’re rolling out the solution one step at a time since there’s a lot to absorb.
Yes, this also hasn’t been done yet but is coming. For now we would tell 3rd party developers to play around if they want but to hold off on building apps until we release our tools
Yes, will do relatively soon
We’ve already built a lot of this and are going to start releasing things quite soon! Some things like the framework for 3rd party developers is in active development. To clarify when you ask “how much of this is implemented on Geo”, for terminology, I believe the knowledge graph should be on The Graph. That includes the data structures (hence why GRC-20 is a Graph Request for Comment) as well as the governance. Ultimately I want Geo to just be the generalized user interface for browsing and contributing knowledge. Many apps will be able to write to and read from The Graph.
Also, yes, the work @0xThierry and team are doing is awesome. AI Agents will need to have a source of truth for asking questions against and taking actions. A global knowledge graph with public governance would be the perfect substrate.