Moxie's Thoughts on Cryptocurrencies and Web3

Opens February 4th 2022 recovered from Omicron, and says he will read a Moxie Marlinspike blog post together with the audience.

Shouts out Cohen for a Hoskin Tunes gift featuring Logan the lobster, then shares his screen to read the post.

Introduces Moxie Marlinspike, co-author of the Signal protocol and former Whisper Systems CTO, a prominent privacy advocate whose essay My First Impressions of Web3 captures the industry's state.

Reads Moxie saying that despite being a cryptographer he is not drawn to crypto and does not share the generational excitement for moving all of life into an instrumented economy.

Summarizes the web3 thesis that web1 was decentralized, web2 centralized everything into platforms, and web3 will re-decentralize while keeping web2's richness.

Reads Moxie's first axiom that people do not want to run their own servers and never will, even nerds, relating it to how Daedalus nodes are a poor user experience.

Companies that ran servers for you like Google and Facebook succeeded because protocols evolve far slower than platforms, since email is still unencrypted after 30 years while WhatsApp encrypted in a single year.

Highlights the key point that if something is truly decentralized it becomes hard to change and stuck in time, which is exactly his Bitcoin critique since the ecosystem moves fast.

Decentralized protocols becoming hard to change is why the hard fork combinator and governance are critical, since protocols stuck in stasis have limited utility, a big Cardano focus.

Reads that Moxie built two dapps to feel out web3: Autonomous Art, an NFT minting piece whose cost rises over time into a pyramid shape, and First Derivative for NFT derivatives.

The dapps are just normal React websites whose distributedness only refers to state and update-permission logic living on the blockchain, and that crypto ignores the client-server interface.

Reads the crux that all network diagrams treat clients as observers, and blockchains are a network of peers not designed for your mobile device or browser to be one of those peers.

In a mobile world of clients and servers Ethereum even calls servers clients, with billions more clients than servers, so a dapp must reach the blockchain via a remote node.

Reads that companies emerged selling API access to Ethereum nodes, and almost all dapps use just Infura or Alchemy, with even MetaMask usually just calling Infura.

The real crux is these client APIs lack inclusive accountability, since responses are not signed or verifiable client-side, so you just trust the JSON blob Alchemy or Infura returns.

Reads that Moxie made a traditional NFT and found there is no hash commitment for the data at an NFT's URL, so the image cannot be verified as authentic.

Most NFTs just point to a URL on some Apache VPS that anyone owning or compromising the machine can change, with nothing in the spec confirming the correct image.

Reads that Moxie made an NFT showing different images by IP or user agent, looking one way on OpenSea and another in the wallet as a poop emoji, proving many high-priced NFTs could change at any time.

Reads that OpenSea removed his NFT without explanation despite no terms violation, after which it vanished from every crypto wallet on his device.

A non-custodial wallet like MetaMask or Rainbow keeps keys client-side but cannot be a blockchain peer, so it just makes unauthenticated API calls to Etherscan, Infura, and OpenSea, reusing TLS sessions that let those companies link your accounts.

Rainbow is set up the same way, building social features on Firebase rather than the blockchain, so if OpenSea removes your NFT it disappears from your wallet even though it is still on-chain.

Reads that web3 recreated web1's trappings and is consolidating around platforms again like Infura, Coinbase, and Etherscan, while the protocols themselves evolve slowly without built-in governance.

NFT royalties were not specified in the ERC-721 standard and it is too late to change, so OpenSea configures them in web2, meaning centralized platforms outpace the distributed protocols.

Reads what he calls the most important line: once a distributed ecosystem centralizes around a platform for convenience it becomes the worst of both worlds, centralized control yet still mired in time.

Reads Moxie's response to the it-is-early-days refrain, that these technologies tended toward centralization from the very beginning with zero effect on velocity because most participants do not know or care.

Reads that people want only the minimum decentralization for something to exist, and that OpenSea would be faster and cheaper without web3, since accepting a bid cost Moxie 80 to 150 dollars in Ethereum gas fees.

Reads that it is a gold rush where speculators want to spend crypto in ways that support their investment, so the money draws people to OpenSea which iterates on web3 in the web2 space.

Reads the endgame where Coinbase offers NFT access by debit card via dark pools, eliminating smart-contract interaction, until all web3 components are gone and it is just a website for buying JPEGs, web2 times two.

Reads Moxie asking whether the accumulated cryptocurrency is an engine or a leaky bucket, concluding enough money exists to keep the faucets going so this will not be just a blip.

Reads Moxie's view that web3 is not on a trajectory to deliver people from centralized platforms and its privacy is already below the internet's low bar, though nerds enjoy the creativity springing from its clunky constraints.

Reads Moxie's first recommendation to accept that people will not run servers by designing systems that distribute trust without distributing infrastructure using cryptography, noting how little cryptography web3 actually uses.

Reads Moxie's second recommendation to reduce the burden of building software, since projects need huge specialized teams that serve their own interests and distributed systems make it worse.

Praises the write-up as an infosec outsider giving crypto a fair chance and finding its centralized points of failure, which is exactly why they wrote the Mithril paper for inclusive accountability.

Mithril, extended UTXO's local state, and this research aim to get logic into a proof you can verify, while governance lets protocols evolve as fast as centralized software so functionality is not pulled off-chain.

StarkWare's fractal-scaling blog post recursively embeds structures in centralized infrastructure while keeping an underlying root of trust, and that building protocols that stay decentralized over time is extremely hard.

Asks what dapp-development model forces cryptography at the base so users can verify what they see, since if four companies control the whole experience it is no different from Amazon, Rackspace, or DigitalOcean.

Cardano's goalposts are truly decentralized things that keep decentralizing over time, with consequences like Daedalus being a slow server precisely because it is decentralized, a huge engineering challenge.

Warns of Dunning-Kruger in the industry where people judge only by price, but notes Google's high share price is not a decentralized experience, so step back and ask what you are really achieving.

It would be sad if all this yields a slower, more expensive web2 where five companies decide everything and you can be deplatformed, so alliances like the DeFi Alliance and an NFT alliance must set standards preserving decentralization and inclusive accountability, hashing NFT data and legal nexus.

Concludes that standards and protocols must evolve faster, the great challenge of governance, and that combining scalability, interoperability, and sustainability yields an ecosystem without Ethereum's flaws.