In this installment of our Obliquities editorial column, we argue that the social kernels circulating in intelligence media are the equivalent of industrial intermediates flowing between factories.
The F2F links you are building sound a lot like protocols. You have to collaborate with another person to bootstrap protocols that enable your AI factories to work with one another. Initially these F2F relationships and the protocols that enable them will be bespoke. However, I think its safe to assume that these F2F relationships will be able to scale past "strong trust" relationships to bigger groups that have lower amounts of trust, if effective protocols can be designed. MCP exists, I know - but what you're describing here is a whole different animal. What kind of standards and protocols will need to be designed to enable one person's personalized Claude factory to communicate to another's?
Love this. The factory metaphor is definitely generative. It has me asking: should we expect (or facilitate) standardization to develop in the way that factories interact and "plug in" to each other? Or should we expect (and embrace) F2F connections to remain as bespoke as the personalized factories themselves, and thus resemble something more like the relationships between individual humans, of which they are extensions?
I'm also imagining this trend at maturity further challenging traditional boundaries between organizations. If my factory is connected to 5, 10, 100 others, each connection serving a distinct purpose, and each of those factories is also connected to its own web of connections, then when we zoom out and look at the whole graph, we see a kind emergent web of productivity within which it may not make sense to label fixed chunks of it with legal fictions like "company." We might instead have ways of referring to different chunks of it that are more flexible and dynamic based on the subset of processes and outputs relevant for a particular conversation...
Fully on board with this vision of human/AI collaboration. And the F2F language is great.
Had some similar reflections on AI-native art in a collab last year, and this inspires me to think more about the business context.
The language we used was "kernel" and "conversations". I speculated that kernels would be shapes in the latent space, but it now looks like it will just be markdown files.
The F2F links you are building sound a lot like protocols. You have to collaborate with another person to bootstrap protocols that enable your AI factories to work with one another. Initially these F2F relationships and the protocols that enable them will be bespoke. However, I think its safe to assume that these F2F relationships will be able to scale past "strong trust" relationships to bigger groups that have lower amounts of trust, if effective protocols can be designed. MCP exists, I know - but what you're describing here is a whole different animal. What kind of standards and protocols will need to be designed to enable one person's personalized Claude factory to communicate to another's?
Yep
Love this. The factory metaphor is definitely generative. It has me asking: should we expect (or facilitate) standardization to develop in the way that factories interact and "plug in" to each other? Or should we expect (and embrace) F2F connections to remain as bespoke as the personalized factories themselves, and thus resemble something more like the relationships between individual humans, of which they are extensions?
I'm also imagining this trend at maturity further challenging traditional boundaries between organizations. If my factory is connected to 5, 10, 100 others, each connection serving a distinct purpose, and each of those factories is also connected to its own web of connections, then when we zoom out and look at the whole graph, we see a kind emergent web of productivity within which it may not make sense to label fixed chunks of it with legal fictions like "company." We might instead have ways of referring to different chunks of it that are more flexible and dynamic based on the subset of processes and outputs relevant for a particular conversation...
Fully on board with this vision of human/AI collaboration. And the F2F language is great.
Had some similar reflections on AI-native art in a collab last year, and this inspires me to think more about the business context.
The language we used was "kernel" and "conversations". I speculated that kernels would be shapes in the latent space, but it now looks like it will just be markdown files.
https://www.octopusyarn.com/i/173017289/the-art-is-in-the-conversation
Exactly. There's a thinking going on which is basically L2L (have your lab call my lab).