If you want to follow the debate about crypto’s impact on society, which I believe is one of the most important topics in tech today, you better sharpen your Twitter skills – most of the interesting thinking is happening across Twitter’s decidedly chaotic platform. I’ve been using the service for nearly 15 years, and I still find it difficult to bring to heel. When following a complex topic, I find myself back where I started – in a draft blog post, trying to pull it all together.
That’s where I’ve been this past weekend as I watched the response to a thoughtful post from Signal founder Moxie Marlinspike. (And yes, the fact that the Twitter conversation was driven by a blog post is not lost on me…)
Wrote some notes summarizing my first impressions of web3: https://t.co/16MoPROOGc
— Moxie Marlinspike (@moxie) January 7, 2022
For those of you who might not use Marlinspike’s service, Signal is an encrypted messaging platform favored by pretty much everyone in the tech and media world. Marlinspike’s post laid out several shortcomings of the current web3 world, all of it based on his own extensive “tinkering” with things like minting NFTs and building distributed apps, or dapps. It’s worth reading the whole thing, but to summarize, his critique has three key points:
First, while web3 is supposed to be about a world free of centralized services, it turns out most of the well-known web3 platforms (OpenSea, Coinbase) are, in fact, centralized just like web2 (this echoes a criticism brought up earlier in the week by Ben Thompson (sub required, worth it).
Secondly, technical protocols evolve slowly – and protocols are the basis for a lot of web3’s magic. Marlinspike points out that most web1 protocols – like SMTP for mail – are stuck in time and fail to evolve. This is often because the protocols are decentralized – no one is in charge of improving them.
Thirdly, there’s a lot of room for error, mischief, or worse in how many of these services and protocols currently interact – particularly around fundamental issues of trust and privacy, two pillars of web3 philosophy. Marlinspike uses the example of an NFT he created which was banned by OpenSea and subsequently disappeared from his MetaMask wallet to make his point.
If you’re still reading, congrats – that’s a lot and we’ve not yet gotten to the good stuff, which for me is the discussion that’s evolved since Marlinspike’s post. Watching the responses come in felt a lot like reading the early blogosphere – one by one, people I admire built on Marlinspike’s thinking, challenging some of it here, deconstructing other parts there. The tone was respectful, considered – no one reacted as if their religion had been impugned.
The first response I noticed was from Vitalik Buterin, co-creator of Ethereum.
I think the status quo comes from a present-day "missing middle" where we have centralized-but-easy things and decentralized-but-hard things, but new tech (w lots of cryptography!) actually is on the cusp of giving us the best of both worlds.https://t.co/XIx8f2oiVi pic.twitter.com/phTSrGW0OR
— vitalik.eth (@VitalikButerin) January 8, 2022
Buterin challenges Marlinspike’s focus on technical grounds, particularly the term “servers,” and reminds us that there’s still a ton of infrastructure and foundational software work to be done. He points out that 2022 will be a big year for ETH, given its shift from the slower and most costly proof of work to the more nimble and efficient proof of stake.
I then realized I had missed Brian Armstrong’s response, which came a few hours after Marlinspoke’s initial post:
3. I agree there is an overall move toward using platforms. But there is a big difference between using a platform that also owns all the data also (web2) and a platform that is merely a proxy to decentralized data (web3).
— Brian Armstrong (@brian_armstrong) January 8, 2022
Armstrong runs Coinbase, arguably one of the most centralized “web3” companies built so far. His last point is key: There’s a big difference between a company built to control data (Facebook) and one that acts as a useful wrapper for data owned and controlled by the end user. VC Chris Dixon elaborates in a thread the next morning:
How web3 data portability reduces the power of centralized services 🧵
— cdixon.eth (@cdixon) January 8, 2022
Dixon is pointing out a key distinction between web2 and web3 services, regardless of their potentially centralized nature: Ease of data portability. I’ve long argued that any apps or platforms based on leveraging our data should compete on the quality of service they provide, rather than the data they lock in. In 2008, I wrote “It’s time that services on the web compete on more than just the data they aggregate.” This is Dixon’s point in a nutshell: “web3 works like web1 did. There will be centralized services built in web3 — and many will be quite useful — but their economic power and overall control will be limited by the lower switching costs due to data portability.”
The discussion continued later that day with Matt Mullenweg, the CEO of Automattic, the company behind WordPress. WordPress drives more than 40% of the current internet, and Mullenweg has long been a standard bearer for web2’s original philosophy – that of interoperability.
Anyway the most interesting conversation to me is how to make all the new stuff as user-friendly as possible without recreating the same centralized bottlenecks and gatekeepers.
— Matt Mullenweg (@photomatt) January 9, 2022
Mullenweg name checks my former partner Tim O’Reilly, whose seminal “What Is Web 2.0” paper kicked off our Web2Summit conference series and has helped frame my thinking about the Internet for the past 15+ years. Mullenweg’s point is that many original web2 services are entirely consistent with web3 philosophies. That is still true today – whether or not web3 technologies are at the core of it (Mullenweg himself might best be described as “extremely crypto curious.”)
Debate on Marlinspike’s post continued throughout the weekend, and by Sunday, former Dropbox CTO Aditya Agarwal responded elegantly to Marlinspike’s second point, that of protocols.
1/ @moxie wrote a fantastic blog post about his observations on Web3 that really got me thinking about some underlying questions of centralization, protocols, platforms and lastly our ability to predict stuff in this space. 🧵
— Aditya Agarwal (@adityaag) January 9, 2022
Remember that Marlinspike’s criticism of protocols is that they are slow to evolve. Agarwal explains that while this was true of protocols in the early web, it’s not necessarily true in web3 architectures. “…everyone’s mental model of ‘protocols’ is that of current ones like HTTP, SMTP etc. All of those protocols are *stateless*. That has been the accepted (and generally right) model of protocol design. The biggest difference for web3 is that they are stateful protocols. In that sense, I think that pace of protocol evolution isn’t really the right mental model. If the state is generally accessible, then it is much easier to remix and compose. There haven’t been too many instances of such ‘protocols’ which is why it isn’t surprising that all of us are unsure about how to compare this to traditional models.”