Great piece - so much bureaucracy is required for coordination. Handing that over to AI is one of the cases I where I think full automation is truly welcome. I wonder the same for things like representative democracy (do we need politicians?) and money (could AI facilitate a highly coordinated purely barter economy?).
Speaking as a gestational $20/month code-pupa, it is already visible for me to see all of this in action, and the extraordinary difference between the output I'm achieving on my own projects versus the agonisingly slow, institutional world of my job (UK consumer finance, very old-fashioned). Even though I haven't got a fully-set carapace yet, what you are describing explains the situation perfectly.
Your description, one supposes, goes to show that good writing and the edifying use of metaphor, as something requiring an understanding of surprise and taste, is one of the things that will remain in the hands of us humans for a good while yet!
I think institutions could be bottom-up though, no? Wikipedia, burning man and the republic of letters I think all classify as institution as per common defintions:
North: "Institutions are the rules of the game—humanly devised constraints that structure interaction."
Mokyr: "Institutions are socially determined conditional incentives and consequences to actions, parametrically given to individuals and beyond their control."
Hadfield (& Weingast): "Institutions are the processes by which a shared classification of behavior as acceptable or not is generated and enforced."
I want this future to play out so parasitic corporations can die,but the analysis has two big blind spots. First , deep domain knowledge is totally automatable. It's already happening as gig workers with domain knowledge train AI. Techniques like Cognitive Task Analysis can elicit latent knowledge into forms frontier models can act on with sufficient context engineering. Second blind spot is capital equipment. Maybe an exocorporation can design an ASIC, but who's going to pay TSMC millions to build it? How's TSMC going to turn into an exocorporation? Are their multi billion dollar fabs going to dissolve after every production run? Who's going to buy chips and servers from a collective that's going to dissolve rather than stay around and provide support? The pandemic showed us what happens when supply chain commitments aren't met. How will exocorporations make any commitments at all?
Good essay, notwithstanding the overabundance metaphors taken from invertebrate lifeforms (trees and forests are nice, but crustaceans a bit less so).
Have you given thought to what happens to the owners of buildings, buses, machines etc? That is, owners of capital in the old-fashioned sense?
Also, if you'll forgive me one "ackshually": trees and fungi are symbiotic in a forest, it is trees and grasses that compete (trees rely on mycelium for breaking down nutrients in soil whereas grass soils are more reliant upon bacteria for this task—to my understanding while both forest biomes and grassland biomes represent evolutionary stable systems, the forest biome supports a greater biomass and has a higher energy capture than grassland). Not that it's particularly relevant to your metaphor, but that at least is what I was given to understand from my amateur interest in permaculture when I was younger.
Anyway, excellent analysis, it was worth subscribing just for this.
Loved the perspective. As a solo consultant now working with 2 agents (OpenClaw and Hermes agent) each able to spawn their sub agents, this feels very timely. Also quite different from the view Jack Dorsey lately out out about corporations leaving the org chart behind in favor of intelligence weaved into the fabric of the company. Will be cool to see what plays out and how!
Ever since reading "Life After Google", where George Glider uses Ronald Coase's work to predict Google's demise, in 2020 I've been keeping Ronald Coases work in the back of my mind. This is now the third take I've read on the mixing between his ideas and AI.
I think you did a good job comparatively because you talked about trust. A lot of the transaction costs that we place in the market are there to make up for a lack of trust. In some cases the friction acts as collateral, acts as a filter, or exists in legal wording to give weight to counter parties. However you can trust Susanne from the Billing department (until she ate your salad from the fridge).
But I'm not sure that per se transaction costs will in fact collapse. I think how our society will integrate trust is always hard to predict. Also there is an element of obfuscation that occurs with AI. Yes AI can investigate for you to help make verification easier however there also will likely be more to verify. If everyone is using AI to check everyone else who are also using AI to do things then you need N^2 verification actions. What you are talking about only works if a persons ethics shine through in everything they do (ala New Game+). At which point verification can cut right through the bulk.
Can Neo-China stop arriving from the future. I just want to live my life
Great piece - so much bureaucracy is required for coordination. Handing that over to AI is one of the cases I where I think full automation is truly welcome. I wonder the same for things like representative democracy (do we need politicians?) and money (could AI facilitate a highly coordinated purely barter economy?).
Absolutely wonderful post.
Speaking as a gestational $20/month code-pupa, it is already visible for me to see all of this in action, and the extraordinary difference between the output I'm achieving on my own projects versus the agonisingly slow, institutional world of my job (UK consumer finance, very old-fashioned). Even though I haven't got a fully-set carapace yet, what you are describing explains the situation perfectly.
Your description, one supposes, goes to show that good writing and the edifying use of metaphor, as something requiring an understanding of surprise and taste, is one of the things that will remain in the hands of us humans for a good while yet!
I think institutions could be bottom-up though, no? Wikipedia, burning man and the republic of letters I think all classify as institution as per common defintions:
North: "Institutions are the rules of the game—humanly devised constraints that structure interaction."
Mokyr: "Institutions are socially determined conditional incentives and consequences to actions, parametrically given to individuals and beyond their control."
Hadfield (& Weingast): "Institutions are the processes by which a shared classification of behavior as acceptable or not is generated and enforced."
Yes, that's fair. Probably not possible to have a "pure" extitution, especially with a broad definition, there will always be institutional elements.
"Extitution" is foregrounding relationships over formal structures.
I want this future to play out so parasitic corporations can die,but the analysis has two big blind spots. First , deep domain knowledge is totally automatable. It's already happening as gig workers with domain knowledge train AI. Techniques like Cognitive Task Analysis can elicit latent knowledge into forms frontier models can act on with sufficient context engineering. Second blind spot is capital equipment. Maybe an exocorporation can design an ASIC, but who's going to pay TSMC millions to build it? How's TSMC going to turn into an exocorporation? Are their multi billion dollar fabs going to dissolve after every production run? Who's going to buy chips and servers from a collective that's going to dissolve rather than stay around and provide support? The pandemic showed us what happens when supply chain commitments aren't met. How will exocorporations make any commitments at all?
Good essay, notwithstanding the overabundance metaphors taken from invertebrate lifeforms (trees and forests are nice, but crustaceans a bit less so).
Have you given thought to what happens to the owners of buildings, buses, machines etc? That is, owners of capital in the old-fashioned sense?
Also, if you'll forgive me one "ackshually": trees and fungi are symbiotic in a forest, it is trees and grasses that compete (trees rely on mycelium for breaking down nutrients in soil whereas grass soils are more reliant upon bacteria for this task—to my understanding while both forest biomes and grassland biomes represent evolutionary stable systems, the forest biome supports a greater biomass and has a higher energy capture than grassland). Not that it's particularly relevant to your metaphor, but that at least is what I was given to understand from my amateur interest in permaculture when I was younger.
Anyway, excellent analysis, it was worth subscribing just for this.
Loved the perspective. As a solo consultant now working with 2 agents (OpenClaw and Hermes agent) each able to spawn their sub agents, this feels very timely. Also quite different from the view Jack Dorsey lately out out about corporations leaving the org chart behind in favor of intelligence weaved into the fabric of the company. Will be cool to see what plays out and how!
Thank you. Very interesting.
Ever since reading "Life After Google", where George Glider uses Ronald Coase's work to predict Google's demise, in 2020 I've been keeping Ronald Coases work in the back of my mind. This is now the third take I've read on the mixing between his ideas and AI.
I think you did a good job comparatively because you talked about trust. A lot of the transaction costs that we place in the market are there to make up for a lack of trust. In some cases the friction acts as collateral, acts as a filter, or exists in legal wording to give weight to counter parties. However you can trust Susanne from the Billing department (until she ate your salad from the fridge).
But I'm not sure that per se transaction costs will in fact collapse. I think how our society will integrate trust is always hard to predict. Also there is an element of obfuscation that occurs with AI. Yes AI can investigate for you to help make verification easier however there also will likely be more to verify. If everyone is using AI to check everyone else who are also using AI to do things then you need N^2 verification actions. What you are talking about only works if a persons ethics shine through in everything they do (ala New Game+). At which point verification can cut right through the bulk.