Hypermachines: Realtime and Collaborative P2P Search

Modern search systems don't work offline, rely on proprietary indexes, and give users limited interfaces for content discovery. Our earlier work on the Hypercore Protocol produced a collection of data structures and networking modules for building low-latency, secure P2P applications. With this project, we will extend the Hypercore Protocol with a novel mechanism for distributing sandboxed computation, called Hypermachines, that can be combined with the existing data structures in our stack to power a next-generation search system. Hypermachines are deterministic Javascript programs, akin to lightweight smart contracts, that introduce algorithmic transparency and compositionality into our ecosystem. Users can create powerful indexing pipelines that merge their Hypermachine datasets together, yielding a highly-composable, collaborative search engine. By storing indexing logic directly alongside data structures, users can see exactly how indexes are produced, verify that they were produced correctly, and modify them according to their needs. We imagine a future in which Hypermachines power a decentralized marketplace for collaborative, transparent, and fast search engines.