Building secure systems from independent, mutually distrustful parties is an old topic in computer science. But despite its attendant hype and misinformation, today's “blockchain bandwagon” has successfully brought the gospel of decentralization - both a realization of its possibility and an appreciation for its value - to mainstream society. Currently-deployed blockchains, however, are slow, unscalable, weakly consistent, profligate in energy use, and have effectively re-centralized due to market pressures. We will explore ongoing challenges and progress in rethinking blockchain architecture to improve scalability, efficiency, functionality, privacy, and decentralization. We will explore how decentralized building blocks such as collective signatures and scalable distributed randomness enable architecturally modular solutions to challenges such as scalable Byzantine consensus, horizontal sharding, proof-of-stake, and blockchain-managed secrets. Finally, we explore challenges in fairness and democratization in decentralized systems, how “proof-of-personhood” blockchains could enable information forums and anonymous reputation systems resistant to propaganda campaigns, and how democratic cryptocurrencies could offer a permissionless analog of universal basic income.