Mycelium: A Peer-to-Peer Chat System for Metadata-Minimizing; AI Infrastructure is Reshaping Communities and Regulation is Falling Behind 41 views

My STS Paper discusses how data centers have shaped policy and communities, and how the AI infrastructure is evolving around the world, and what can be done to steer the growth before it goes out of control. I argue that the technical momentum of data centers is clear. They have shifted from infrastructure shaped by society to infrastructure that now shapes society, producing through environmental and economic impacts that affect communities around the world. The policy positions presented in the paper include approaches from the U.S. and EU as well as proposals about how the power grid must transition to support the infrastructure boom happening with data centers. The U.S. policies I examine offer a deregulation approach, in which the current administration wants to fastrack approvals by removing guard rails. The EU’s approach is one where the end-goal of data center growth is the same, but they want to maintain environmental and efficiency standards for their infrastructure. My analysis weights two positions as well as other proposals such as the Twin Transitions Philosophy to provide recommendations to the United States government on what direction they should take with data center policy. My Technical Project, Mycelium, is a decentralized peer-to-peer chat application that my capstone partner and I developed to address a limitation with most secure messaging applications. Most secure chat applications like Signal advertise end-to-end encryption to their users, but message metadata is still being stored on central servers. The centralized servers pose a privacy risk because the data they hold can be exposed to subpoenas and data breaches. To address this, my teammate and I built a fully peer-to-peer desktop chat application. We developed the backend in Go using the libp2p library and packaged the application with Wails, which exposes the Go backend to a JavaScript and HTML/CSS frontend. The application uses 3 bootstrap nodes that we host on AWS EC2 to help peers discover others in the same chat room. Once two nodes are connected, messages are sent directly between them, and message history is stored in a local LevelDB instance packaged with the application. We evaluated the system through automated connection benchmarks, scripted functional walkthroughs, and security validation including packet captures and bootstrap-shutdown tests. Mycelium achieved a 100% success rate when connecting to other peers, but the main limitation right now is the connection latency with average times ranging between roughtly 25 to 72 seconds, which is the major roadblock for production ready use. Both projects respond to the same underlying phenomenon, which is the rise of data and compute economy where information and processing power have become a new form of currency. The technological momentum of this economy has reshaped software and hardware infrastructure alike, often faster than communities or users can respond to it. The costs of this new economy fall unevenly on communities that host the physical infrastructure and on users whose data accumulates in systems they cannot see or control. My STS Paper and techological project address this shift at two different scales. The STS paper engages with the physical layer, which includes the data centers that produce and process information at scale, and it asks how policy can shape that growth to limit the harm to the communities around the infrastructure. The technical project addresses the application layer by offering an alternative architecture that limits how much user data reaches centralized servers in chat applications. Together these two projects suggest that limiting the momentum of data and compute requires intervention at multiple levels, through policy that shapes where infrastructure can grow and at what cost, and through software design choices that reduce how much data that infrastructure holds in the first place.

BS (Bachelor of Science)

Data Center; AI Infrastructure; Decentralized Communication; AI Regulation

School of Engineering and Applied Science Bachelor of Science in Computer Science Technical Advisor: Mark Sherriff STS Advisor: Joshua Earle Technical Team Members: Sri Kapa

English

All rights reserved by the author (no additional license for public reuse)

2026-05-11