a16z Podcast: Five Open Problems Toward Building a Blockchain Computer

Episode Summary

Executive Summary: Frank Chen and Ali Yahya frame crypto as a new computing paradigm centered on decentralized trust rather than mere payments. They break down the blockchain computer into three open scalability problems—compute, storage, and networking—and argue that advances in consensus, CDNs, proofs of retrievability, and identity/governance could enable new classes of applications, especially where trust is the bottleneck to adoption.

Main Topics: Crypto as a new computing paradigm (Priority: 5/5): Ali argues blockchain is best understood as a decentralized computer that enables trust-minimized computation, not just a payment network. The core innovation is decentralizing human control over systems. Scalability of distributed compute (Priority: 5/5): The discussion separates compute scalability into throughput (instructions per second), latency to finality, and cost per instruction, emphasizing that these are still open problems limiting real applications. Proof of work vs. proof of stake (Priority: 5/5): They explain why proof of work is expensive and energy-intensive, and how proof of stake replaces external resource consumption with token ownership to reduce cost while preserving trust properties. Distributed storage markets (Priority: 4/5): The conversation explores Filecoin, Sia, and Storj as examples of decentralized storage networks that rely on cryptographic proofs of retrievability to ensure providers actually store user data. Decentralized networking and CDNs (Priority: 4/5): They discuss blockchain distribution networks, mesh networking, and incentivized packet-forwarding as efforts to reduce propagation delay and decentralize communication infrastructure. Trust, identity, and reputation in Web3 (Priority: 5/5): Ali contrasts Web2 identity/reputation systems with crypto’s pseudonymous, incentive-based trust model, and highlights identity/reputation as an unsolved area that could make decentralized systems more usable. Governance and protocol evolution (Priority: 4/5): The episode closes on governance: how decentralized systems update rules, coordinate stakeholders, and avoid the same power-concentration issues seen in centralized tech platforms.

Key Arguments: Crypto’s defining feature is decentralization of human control, not geographic distribution or simply a new database. The right scalability metrics for blockchains-as-computers are throughput (instructions per second), latency to finality, and cost per instruction. Current blockchain limits are heavily influenced by network propagation delay and message size, creating a trade-off between performance and decentralization. Proof of work is expensive because it forces participants to spend real-world electricity/computation to enter consensus; proof of stake reduces this by using owned tokens as the participation resource. Decentralized storage is valuable primarily when trustlessness matters; it is not meant to beat Amazon or Google on raw price for ordinary photos/backups. Proofs of retrievability are a core cryptographic challenge for storage networks because they must verify that a provider really retains user data. Networking is a hard but necessary layer to decentralize; incentivized routers, mesh networks, and block propagation systems can improve censorship resistance and performance. The most promising crypto applications are those where trust is the bottleneck to scaling: money, lending, derivatives, exchanges, gaming assets, and identity-linked marketplaces. Open-source development and composability may create unusually fast, combinatorial innovation across apps built on the same decentralized fabric. Identity and reputation could transform crypto from a series of one-off prisoner’s dilemmas into iterated relationships with stronger cooperation incentives. Governance remains unsolved: decentralized systems need ways to change over time without recreating centralized control or becoming stuck. Crypto will not replace centralized systems everywhere; centralized architectures will remain superior for many cost-sensitive, non-trust-critical use cases.

Data Points: Bitcoin finality wait time: about 60 minutes - Ali cites this as the approximate period needed to be reasonably sure a proof-of-work transaction will not be reverted. Bitcoin block time: 10 minutes - Used as an example of the cadence of blocks in the Bitcoin network when discussing block size and propagation delay. Block size example: 1 MB to 2 MB - Referenced in the Bitcoin block size debate to illustrate the throughput vs. propagation-delay trade-off. Ownership-to-voting share in proof of stake: 2% tokens → ~2% say - Ali explains proof of stake as proportional participation based on token ownership. Ethereum emergence: 2014 - Cited as the launch period of Ethereum, helping frame how early the broader blockchain-as-computer idea still is. Time since Ethereum thinking took hold: 4–5 years - Ali says it has only been a few years since people widely began thinking of blockchains as computers rather than payment systems.

Pivotal Quotes: "When people talk about decentralization in the world of crypto, they mean decentralization of human control." — Ali Yahya: Definition of what decentralization means in the blockchain-computer framing. "The right metric is not really transactions per second, it's really instructions per second." — Ali Yahya: Explaining why blockchain scalability should be evaluated as general computation, not just payments. "Decentralized systems will always be more expensive and less performant than centralized ones." — Ali Yahya: A key caveat: crypto wins where trust matters, not where raw efficiency dominates.

Implications: For engineers, crypto is presented as a frontier of unsolved distributed-systems problems. The biggest opportunities lie in infrastructure, identity, governance, and trust-heavy apps—not simple Web2 replacements.