mirror of
https://github.com/lux-apps/papers.git
synced 2026-11-02 11:39:52 +00:00
zeekay
abf99d05fc
Restructured from flat .tex/.pdf files into per-paper subdirectories. Added Antje Worring as co-author on all papers. Synced PDFs to pdfs/ for download site.
20 lines
2.2 KiB
TeX
20 lines
2.2 KiB
TeX
\section{Related Work}
|
|
\label{sec:related}
|
|
|
|
\textbf{AutoGPT}~\cite{autogpt2023} pioneered autonomous agent loops with persistent file storage and web access. However, agents have no cryptographic identity, no capability scoping, and no economic model. Any agent can access any file in its workspace.
|
|
|
|
\textbf{LangChain}~\cite{langchain2023} provides a composable framework for building agent pipelines with tool use, retrieval-augmented generation, and chain-of-thought prompting. Agents are stateless by design---persistence requires external infrastructure, and there is no native identity or economic layer.
|
|
|
|
\textbf{CrewAI}~\cite{crewai2024} introduces role-based multi-agent coordination with hierarchical delegation. Roles are application-level labels, not cryptographically enforced capabilities. There is no on-chain settlement or private memory.
|
|
|
|
\textbf{Eliza (ai16z)}~\cite{ai16z2024eliza} adds on-chain token interaction to conversational agents, enabling agents to hold and transfer tokens. However, agents share a single identity model, lack fine-grained capability scoping, and have no private inference pipeline.
|
|
|
|
\textbf{Virtuals Protocol}~\cite{virtuals2024} tokenizes agent personas as NFTs and creates revenue-sharing models for agent creators. The focus is on economic coordination of agent personas, not on the sovereignty primitives (vault, capabilities, private memory) that we address.
|
|
|
|
\textbf{ARC (AI Risk Council)}~\cite{arc2024} proposes governance frameworks for autonomous AI agents but does not specify a technical architecture for agent sovereignty. Our work provides the concrete protocol-level primitives that such governance frameworks require.
|
|
|
|
\textbf{OLAS (Autonolas)}~\cite{olas2023} defines autonomous agent services with on-chain registration and economic incentives. The architecture focuses on service coordination rather than individual agent sovereignty---agents in OLAS share service-level identity rather than possessing individual capsules.
|
|
|
|
Agent Capsules differ from all prior work in that sovereignty is a protocol-level primitive, not an application-layer add-on. Identity, capabilities, economics, privacy, and accountability are enforced by the trust kernel, not by the agent's own code.
|
|
|