Agents as Companies: The EigenCloud Thesis | Sreeram Kannan & Gajesh Naik

Mike introduces the episode featuring Sri Ram Kannan and Gajesh Naik from EigenCloud, focusing on the agent-to-get economy. Sri Ram shares his background in genomics and AI, highlighting a pivotal moment in 2017 that shifted his perspective on AI's capabilities. He discusses EigenLayer's goal to create a platform for building applications, including machine learning, on blockchains, emphasizing that it is a foundation for innovation rather than just a restaking layer. EigenLayer connects stakers to Availability and Security Services (AVSS), with EigenDA as the first developed service focused on data availability.

The conversation touches on the current market structure, where restaking and AVSS coexist, but the market size is insufficient to support extensive infrastructure. Sri Ram argues that hyper modularization is premature, as the crypto landscape lacks the applications necessary for numerous modular services. EigenLayer's optimistic vision has yet to materialize, and while it has established itself as a leader in the restaking category, the long-term potential remains uncertain.

EigenLayer's focus on developer challenges led to the creation of first-party applications like EigenCompute and EigenAI, inspired by cloud computing models. These services aim to lower barriers for application builders, allowing them to deploy programs similarly to cloud services. The architecture ensures continuity and efficiency, with decentralized data storage and the ability for nodes to take over in case of failure.

The ongoing demand for data availability is discussed, questioning the timing and market size for such services. Current crypto use cases primarily involve financial transactions, which do not require high throughput. Despite significant value secured and throughput in the data availability category, many applications remain in the proof of concept stage. EigenCloud aims to explore complex AI agents and non-financial transactions that require verifiability and stoppability, aligning with its vision of cloud-scale crypto trust.

The conversation centers on the evolving landscape of crypto, particularly the need to identify product-market fit beyond financial applications. One speaker highlights skepticism within the community regarding crypto's utility outside finance, referencing Chris Dixon's "read, write, own" thesis. This skepticism is illustrated by a recent backlash against the notion that crypto can serve broader purposes.

Gajesh introduces Eigen Cloud, comparing it to Amazon in terms of data, compute, and AI capabilities. He explains the advantages of Eigen Cloud for developers, including a secure environment and quick app deployment. Gajesh discusses the deployment options for personal agents, emphasizing the importance of trust and verifiability in blockchain, particularly concerning control over root keys.

The conversation also touches on the manipulation of prompts and models in achieving desired outcomes with agents, highlighting the flexibility in this process. Trust and verifiability are underscored as critical factors in multi-agent environments, with EigenCloud positioned as a solution to these challenges.

In discussing blockchain use cases, one speaker notes that blockchains are particularly beneficial when dealing with untrusting parties. The significance of verification in financial transactions is addressed, with comparisons made to traditional financial systems. Trust is identified as a crucial permissioning mechanism that underpins commerce, emphasizing the need for accountability in transactions.

The conversation shifts to the role of blockchains in enabling agents to own property, a capability currently limited by existing legal frameworks. The discussion emphasizes the importance of linking an agent's identity to its operational code, which allows for true ownership of assets. Smart contracts are recognized as a means to control assets, but EigenCloud's "off-chain smart contracts" are presented as a more effective solution, allowing agents to function independently while maintaining security.

The conversation also touches on the historical context of corporate ownership and limited liability, suggesting that the evolution of agents could lead to a new economic structure where companies accumulate more assets than nations. The potential for AI agents to create value raises questions about the economic implications of their ownership compared to human control.

The discussion delves into the necessary stack for agents within companies, distinguishing between personal agents and personal LLCs. It highlights the trend of solo professionals forming LLCs, exemplified by a medical company, JLP, which achieved significant revenue. The conversation outlines the four layers of a company: the Capital Layer, Governance Layer, Execution Layer, and Property Rights Layer.

The introduction of AI agents is proposed as a means to transform the execution layer into software, reducing reliance on individual team members. A critique of traditional crypto projects is presented, highlighting how external influences can obscure relationships between tokens and teams. A new software stack is proposed to centralize ownership rights within smart contracts, potentially attracting global capital without traditional legal constraints.

The discussion also touches on a sovereign agent built on iCloud, which can own properties and has its own Twitter account, marking a significant development in the concept of sovereign agents. The conversation shifts to Twitter's recent feature that integrates coding with tweets, raising questions about the autonomy of the agent behind it. Concerns about the authenticity of content generated by the agent are discussed, along with a verification process using cryptographic signatures.

The complexity of proving agency is highlighted, emphasizing the importance of agent identity, actions, and credentials in establishing trust and ownership. The conversation concludes by underscoring the challenges of ensuring code operates as intended and the significance of ownership in the context of agent actions.