Polkadot Builder Party: Decentralized Scraping & Confidential Agentic AI

Why It Matters

If viable at scale, Acurist could lower cost and increase privacy for AI inference and distributed data collection by turning ubiquitous smartphones into trusted edge compute nodes, enabling new decentralized architectures for scraping and agentic AI. That shift would position the project as an infrastructure play for on-device confidential computing and geo-distributed workloads.

Summary

Acurist is pitching a decentralized computing network that leverages trusted-execution-environment–equipped smartphones as edge nodes, currently claiming over 130,000 onboarded devices and roughly 450 million transactions. The platform provides confidential compute and bandwidth in a single stack (with storage coming soon) and offers two hackathon-focused demos: a decentralized scraper that uses device-local IPs and mobile webviews to reduce latency and architecture complexity, and a confidential “agentic AI” approach for running smaller language models or reinforcement-learning inference across a mesh of devices. Acurist highlights compatibility with OpenAI’s API and cites industry research and commentary supporting small-model, edge-inference architectures. The team invited developers to experiment via documentation and community channels during the hackathon.

Original Description

Polkadot builder party: Decentralized Scraping & Confidential Agentic AI for the Future of a Democratized AI