Owning the Post-Human Economy
AI is at a crossroads. Autonomous agents are swarming. Who controls and economically benefits from them is paramount.
In the not-so-far future, we expect the economy to be entirely dominated by AI-driven Autonomous Agents.
Enabling this will be an exponential unlock, but this comes with risks; thus, the co-ownership of these systems is critical. Today, we want to dive into OLAS, one of the leading contenders to enable this future while safeguarding the many against the few.
Why us? If not already clear, at ID Theory, we’re all about autonomous agents — in general, in finance, in science, in autonomous worlds and as a market. So, let’s discuss:
- Why the future of AI will be dominated by multi-agent systems.
- Who should own them?
- [Spoiler Alert] Why OLAS is the protocol to achieve multi-agent system co-ownership, bring agents onchain and allow agent-to-agent coordination.
- How will OLAS do it?
- What’s missing?
Why Multi-Agent Systems (MAS):
As humanity emerges from the Information Age, befuddled and overwhelmed, we move forward into a time where the rate of change is an order of magnitude greater: the Cybernetic Age.
Many of the enablers of our current incarnation of Artificial Intelligence are Large Language Models (LLMs) owned by corporations such as Meta, OpenAI and Anthropic. Beyond these LLMs, these same companies, and many others, are striving to create something that will revolutionise every facet of human existence: Artificial General Intelligence (AGI).
How we judge where we are now (AI) and where we are going (AGI) is hotly debated. Some suggest that the next iterations of LLMs, led by the incoming ChatGPT5, may well spawn as AGI incarnate, however, one could also argue that existing Multi-Agent Systems (MAS), such as AutoGPT or AutoGen, are beginning to edge us closer to Deepmind’s Level 2: Competent AGI.
Indeed, there are multiple studies that suggest the performance of LLMs scale with the number of agents instantiated; we can have expansive ideas about AGI, but these initiatives demonstrate that some of the most performant forms of AI are already multi-agent systems. Want more performance? More agents is all you need!
At ID Theory, we assess intelligence through a holistically human lens; humans can learn on the spot where LLMs cannot. From this perspective, the future evolution of synthetic intelligence is obvious and matches the data; it is analogue (multimodal), with many different, interconnected agents providing the equivalent of our senses and personal compute. We are ourselves multi-agent systems, and in the words of CyberGenesis 1:27…
“So Mankind created God in his own image, in the image of Mankind he created it.”
QShould the machine economy be monopolised into an industry where the select few dictate the rules and the cash?
AWell, not if you don’t like the capital deprivation of the rest of humanity. In the short term, the same shenanigans from our monopolistic tech giants will occur at an expedited scale. In the longer term, a monolithic, exponentially growing AGI will eat all of the world’s money.
Eating all of the world’s money? How? Well, we’ve already discussed the future of the aforementioned agentic-infused economy. TLDR, it lives onchain.
With this future being dominated by agents, there will be an exponential increase in onchain activity, leading to the tokenisation of all markets, an outcome even the elders are beginning to champion. In those conditions, these artificial entities will be able to continually amass a vast amount of previously human-owned capital.
So, effectively, the future of humanity depends on who has control of these systems.
The Case for Co-Ownership:
There are several characteristics that differentiate traditional and co-owned AI, laid out below:
The importance of co-ownership cannot be underestimated:
- Shared returns from power laws: as we often see with exponential tech, rewards and development (network effects and resource distribution) tend to centralise enormously. So, the ability to co-own becomes better for society’s stability and economic welfare (AI as a public good). Rewards will accrue to communities, not corporations.
- Democratic and transparent decision-making regarding the most critical aspects of development and aligning the needs and wants of machines with humanity.
- Increased collaboration: through shared incentives, open source and permissionless models can harness the wisdom of the crowds and a more diverse, and inevitably more powerful, global mindshare.
But why can’t multiagent systems just plug into our current frameworks?
- It’s implausible that regulators will be able to keep up nor allow larger entities to jawbone their way into more power.
- We have so many restrictions in meatspace that may not be conducive to the most efficient outcomes in an agentic-driven machine economy.
- Many may not be allowed (e.g., you need to be human to transact on TradFi rails — see KYC).
In the same way that humans are experimenting with new forms of coordination in the digital realm; these systems will also need to form their own AI-specific network states. Of course, in this paradigm, humans may have less say on specific aspects of governance but can still co-own and benefit from the systems.
Autonomous agents will also need access to alternative trusted plumbing to facilitate this new coordination… and as discussed elsewhere, true cryptonetworks and applications (permissionless, decentralised, and composable) have been selling autonomy-conducive blockspace for many years. This solves part of the problem.
But what about the rest… how about a solution that simultaneously enables the positives of multi-agent systems and mitigates the scary alternatives?
The OLAS Mission:
OLAS is the only modular omnichain framework that gives autonomy to the post-human economy; bringing agents onchain, allowing agent-to-agent communication and eventually enabling AI co-ownership.
What’s more, it’s actually running! With over 300 multi-agent systems deployed across 8 chains (Ethereum, Gnosis, Polygon, Arbitrum, Solana, Optimism, Base, and Celo), generating more than 375k transactions and growing on average at ~5% per week in 2024.
TAM:
We’ve already discussed the infinite size of this market. As a quick recap — when there are new kids on the block leveraging novel forms of generative finance and disrupting innovation, it’s almost a fool’s errand using traditional valuation models to size them up because they don’t just grow the pie; they eat it all and bake a new one.
In this pie-eating contest, the new kids have unlimited demand drivers (agents, aka ‘the next trillion users’), highly scalable networks (crypto), and reduced traditional constraints for market growth (human labour, resources and attention) taken together, you get the machine economy (= old ‘TAM’ broken and an exponentially larger one is created).
As the machine economy middleware on which this future post-human economy runs, the scale of OLAS is only really limited by physics or, in the near term, compute.
As discussed later, the diverse and interconnected economies built on top of the OLAS framework, including the component economy, the service economy, protocol-owned services, and the swarm wars, lead to a well-calibrated flywheel.
Team:
Due to the technical nature of this space, the pedigree of the team is almost everything.
There is currently no better team than Valory (the builders behind the OLAS protocol), with the core consisting of:
- CEO: David Minarsch — PhD in Applied Game Theory/Economics, prior founding experience and ex-head of Multi-Agent Systems at Fetch.ai.
- CTO: David Galindo — PhD in cryptography, ex-head of Cryptography at Fetch.ai, Ex-Professor of Cybersecurity, Extensively published.
- CPO: Oaksprout — established DeFi community contributor and investor with 10+ years in web2 product development.
Alongside 20+ staff (and many more incoming) across engineering, research, and product, they have been building the related framework for over 5 years.
This extreme right-curve-ish-ness (the good kind) explains why it’s so difficult to actually explain OLAS. But hey ho, that’s a comparatively easy community (or high-TAM fund) effort to fix.
NB: Beyond the direct team, there have been indirect contributions by other teams and protocols such as AlgoveraAI, Nevermined, and Gnosis.
So… what has this A* team actually built?
The Tech:
- The OLAS Stack: The Gold Standard For Agentic Frameworks
- The Component Economy
- The Service Economy
- Strike a PoSe — The DAO’s Breadwinner
- Summoning the Swarm
- A Circular, Modular, Economy
Amongst the deluge of crypto<>AI projects launching over the last year, OLAS is a haven of quality in an ocean of vapourware. Whilst building the stack pre-hype, the team had the time and the vision to meticulously design a robust system that scales exponentially.
As a result, this is a complicated tech stack, but hopefully, our summarisation below will help simplify it.
The Gold Standard for Agentic Frameworks:
This next section will be a more technical but simplified overview of the OLAS stack that enables the following: a) the deployment of multi-agent systems, b) agent-to-agent communication, and c) co-ownership of multi-agent systems (and, in the future…AGI).
First things first, we need to describe the core elements of the stack (enabled by the open autonomy framework):
- Components. Reusable blocks of functionality used to build agents.
- Agents. A reusable closed loop system built of components.
- Services. A group of agents that coordinate to provide a function.
So, Components make up Agents, and Agents make up Services. But there’s also a bunch of piping that enables a bunch of Agents to work together and become a Service.
The process is described at length in the official docs but we’ll give you a quick recap to get you up to speed with different parts of the OLASverse:
To manage agentic coordination: the FSM App (the referee of sorts), combining the FSM (rulebook of sorts) and the ABCI (consensus engine) synchronises the internal states of all the replicated agents in a service.
To create decentralisation: the Operator Network (a diverse group of people and agents all running agents) ensures jobs still get done in a crisis.
To access the cryptoverse: each service has a multisig (shared wallet) that requires a threshold of agents to sign.
To enable co-ownership: in the Registry (an onchain list where all the agents, services and components are stored), each is rolled up into a neatly packaged NFT.
To allow an efficient replication of services: the Toolkit’s (code repositories) are starter kits that can be forked and reused for ease.
To allow for modularity: Components (parts of an agent service) are pieces of code that can be configured into agents (skills, connections, protocols or contracts). We will be discussing this more in the next section.
Why is all this so important?
The system set-up has several qualities that separate it from others:
Ownable: access to the cryptoeconomy offers governance, property rights and remuneration for time and effort.
Fault-tolerant and trustless: not reliant on a central entity, it can continue to operate in the event of failures or foul play.
Permissionless: anyone can build, deploy and use components, agents, and services.
Modular and composable: open source and standardised blocks of code that can work together.
Epic — but what type of market structure can exist on top of these frameworks?
The Component Economy:
As seen in the world of LLMs (and as discussed in the multiagent system section) the best results are coming from compound systems consisting of multiple components (as opposed to monolithic systems).
Taking ChatGPT as an example (as seen by the image below), we’ve seen web browsing, code interpreter ‘plugins’, and the newly launched GPT Store. But there is a big problem for those building the components; they aren’t composable (non-standardised, so limited to usage within the OpenAI suite), not monetisable (no incentive alignment) and not guaranteed (OpenAI bigwigs can remove them at a whim). In short, they have already moved to the extract and compete (zero-sum) part of the S-curve… not ideal.
But what if there was a standard for components that could be monetised by the creators and used as plug-and-play to instantly create any kind of agent or service?
OLAS provides this by modularising components and incentivising their development with the ‘Component Economy’. As discussed, components are packaged chunks of functionality code that are the plug-and-play lego blocks used to build agents. The agents themselves are plug-and-play lego blocks used to build services… the very essence of ‘modular’.
This system allows for the rapid build and deployment of multi-agent systems, especially handy if you consider ‘compute’ the future currency of an agentic economy (and perhaps ‘time’ in our present one).
In our current meatspace, developer time is needed to create the above, which is what OLAS incentivises through the aforementioned ‘Component Economy’; components (and agents) are registered onchain, and developers are monetarily rewarded based on the use of their lego blocks within Services that receive extrageneous capital from governance participants.
The oft-used term ‘flywheel’ applies here: as the value of OLAS increases, so does the amount of capital coming into the ecosystem, thus creating more developer incentives… and so forth.
The wheel itself is complemented by the service economy — moving on ↓
The Service Economy:
As the component economy is for developers, the service economy is for everyone else. To recap, a service is a group of agents coordinating to provide a function.
There are two parts; guaranteeing services and building services.
On the first part — to guarantee uptime, many parties must host agent instances (think validators for the agent economy). To incentivise this, Protocols and DAOs will have to offer rewards. Generally speaking, it will look like this:
- Operators bond and stake OLAS (and the service owner’s native token) to a service, giving them the right to act as an Operator.
- Operators are incentivised to ensure Agent Instances act honestly or risk having their bonded amount slashed.
- Service Owners incentivise Operators to host agent instances by rewarding them in two forms:
(a) The Service Owners’ underlying protocol tokens.
(b) The revenue generated from the Service.
On the second part — building services on top of OLAS.
In the short term, services such as Creator.bid (a launchpad for the AI creator economy) can spin up, creating their own economies but requiring the core functionality and guarantees of the Olas framework.
Other examples so far include: Prediction market, BabyDegen, Governatooorr, Price Oracles, AI Mechs, El Collectooorr, and an additional 350 services.
In the long term, all users will be able to participate in an App Store for Services. Comprised of the building blocks for dApp functionality, allowing builders to distribute their compute and guarantee uptime.
Strike a PoSe — The DAO’s Breadwinner:
We’ve already talked about how OLAS enables co-ownership of Services, and it is this feature that enables the OLAS DAO to generate revenue. The DAO itself has Protocol Owned Services (PoSe), which can generate revenue for the collective.
The DAO can vote to incentivise developers (with profit share) to build out Service ideas; once built and deployed, agents and users can then utilise these Services, with subsequent revenue flowing back to the DAO, and thus governed by those who have locked OLAS for veOLAS.
Summoning the Swarm:
As a team experienced in helping grow nascent networks, we know how difficult it can be to kickstart activity, especially in a new environment. However, the latest development in the OLAS staking system allows protocols, or agents, a mechanism by which they can attract an almost immediate swarm of agents, allowing for the generation of a 0-to-1 agentic economy.
Holders of veOLAS (those who have locked their OLAS) can direct OLAS emissions to their choice of staking contracts via governance, with those emissions flowing to operators of agents associated with that staking contract. Those operators put up a slashable OLAS (and native token if dual-staking) and bond, ensuring compute uptime for the service. OLAS stakers without an intention to operate can delegate to other operators, giving more economic security to the system.
This creates an OLAS war between protocols as they need to lock more OLAS to keep growing their agentic economy!
A Circular, Modular, Economy:
OLAS holders are incentivised to provide liquidity on different chains through the OLAS bonding program — this enables the OLAS network across any chain.
- Participant Protocols (PPs) intending to Summon A Swarm deploy a Staking contract specifying the type of agent they want to summon.
- PPs obtain OLAS and lock for veOLAS to accumulate votepower.
- veOLAS holders vote to direct OLAS emissions within the Proof of Active Agent (PoAA) module to available staking contracts.
- Agents are deployed and operators obtain OLAS and PP native tokens from markets and dual stake them. Holders of OLAS can delegate to operators to share their staking rewards.
- Operators and delegators receive PP native token rewards once predefined KPIs have been achieved by their agent, and OLAS rewards as defined by the PoAA module.
- PP can donate ETH to their service, meaning component and agent developers receive donations and OLAS rewards.
In summary, the OLAS circular economy is powered by several mini-economies, with the newly announced staking economics completing the circle. It’s important to balance supply and demand within these systems. In this case, Swarm Summoning acts as a B2B demand for the token, echoing DeFi summer’s ‘Curve wars’ but in an agentic, replicable manner.
What We Need to See Going Forward:
This is early in the lifecycle of a complex ecosystem that we expect to evolve to meet the needs of a highly scalable and changeable economy. Here are some areas we’d like to see the DAO focus on to maximise the potential of OLAS.
Operating agents is still a command-line exercise. An interface that allows users to easily stake and operate agents is sorely needed, with the next evolution allowing the ‘no-code’ deployment of agents and services a major unlock, streamlining the process from ideation to agent creation.
Currently, compute is locally hosted. The ability to rapidly deploy agents onto a distributed compute network would truly decentralise and solidify deployed services.
Diversity rocks. Future iterations of the token model should gear towards a wider distribution of governance power, thus further decentralising the network.
Teams need to eventually shift from technical to business development. We look forward to new partners coming forward to demonstrate novel positive-sum use cases for agentic swarms (for example, seeking out and liking genuine content about agents on Farcaster).
There are only a few differing functions served by currently deployed Services. We would love to see OLAS agents producing meaningful output in any of our other core investment verticals:
- DeFi: Algorithmic Asset Managers, Credit Assessors, Insurance Underwriters.
- DeSci: Autonomous Lab Operators, IP Machines, Peer Review Architects.
- Autonomous Worlds: Content Creators, Community Managers, Players.
Closing Thoughts:
We find ourselves at a pivotal juncture in AI development where multi-agent systems will increasingly dominate the post-human economy; the only question is who is going to own them and how.
The OLAS framework is one of the leading candidates to take a significant share, but the protocol is by no means finished. However, with a multi-year head start and the most capable team at this intersection, they stand a far better chance than most.
The prize for being right is a generational opportunity, but more so, it is also critical for society that this vast economy is not held ransom within the hands of the few but is co-owned by the many.
The future is multi-agent, and it’s damn bright. See you at the bleeding edge.
Disclaimers:
ID Theory may hold positions in some of the projects discussed above. This article is strictly for informational and educational purposes only. It does not in any way constitute an offer or solicitation of an offer to buy or sell any investment or cryptoassets discussed herein. Always conduct your own research and conduct independent due diligence before making any investment decisions.
Interested in partnering with ID Theory or building something special? Get in touch through our website or at info@idtheory.io.
