The Failure of Web 3 And What Actually Works
The Web3 movement of “owning your data” largely failed. Though there are some loud passionate exceptions, users simply do not care about privacy or “owning their data” enough to sign up or pay for elaborate blockchain applications. Despite the Brave browser being an extremely successful consumer product, for example, its Basic Attention Token has largely failed to reach a meaningful market cap or outperform BTC or ETH. Friendtech was the most recent incarnation, and briefly reached a high market cap before all the users dumped their coins.
At the end of the day, UX determines adoption of consumer apps - and users do not like clicking across multiple steps. At the advertising company I co-founded, which was acquired by Omnicom and ran $100s of millions of ads for Fortune 500 companies, it became clear that each click between a user and a purchase dropped Return on Ad Spend by 30-40%. You add bridging to Arbitrum, or Base as part of your user onboarding and kiss your growth goodbye.
In crypto, users have shown they’re willing to jump through hoops (VPNs, multiple clicks and bridging) to place high value trades - as evidenced by the growth of Phantom Wallet, Polymarket, Hyperliquid and Binance - but they’ve not shown a willingness to use blockchains for data sovereignty, identity, or privacy at any meaningful scale
Crypto is a microcosm of broader society, where this trend holds true across most tech platforms
While every once in a while you hear some horror stories about a creator getting de-monetized by Youtube, the bigger story is being unable to go to the gym without seeing endless influencers shooting selfie videos and onlyfans ‘creators’ making millions per year advertising on Tik Tok and Instagram reels. And - despite alarming trends under the Biden administration, and what seemed like censorship of Donald Trump’s political campaign - he still managed to win the US popular vote.
Users can breathe a brief sigh of relief about data sovereignty.
AI Agents are an Existential Threat to Our Creator Economy
But the equation is changing for five reasons.
- The proliferation of realistic AI video generation models that can render real time photorealistic avatars
- AI agents who - as of October 2024 - have viable “Computer Control” protocols launched by Anthropic, with open source competitors fast behind
- Increasingly growing software stacks such as the Eliza framework that allow AI agents to interact with virtually every aspect of the digital world (hooks for Roblox, X, Youtube, Instagram and more)
- Even large companies have completely failed to protect their IP from being added to “the training data”. For example, ChatGPT can confidently tell you about the detailed plot of all 54 copyrighted Games Workshop Warhammer books as long as you ask politely for a response in Json format.
- Models are getting much smarter and better at creating content that isn’t “cringe” - and resonates with multiple audiences in different languages. When you combine this with the capability of generating streaming video in any language, as well as personalizing messages to each user (via chats) it’s a safe bet that AI agents will create the majority of engaging online content quite soon
As a result – starting in the middle of 2025, the story will stop being “Man I’m too lazy to sign up for that complex protocol that owns my data” to “If I don’t sign up for some sort of system to license my work or have an economic model, I won’t be able to make money as a creator”.
However - even if this is largely true right now, the user interfaces of all Web3 products is still so complex that they’re not seeing traction going up.
Web 4 Is the Solution to this Existential Issue
Meet Web 4. The theory of Web 4 is that AI agents will manage the complex relationship between users and AI models which use blockchains as a data sharing, dispute resolution and governance mechanism.
Let’s create an precise definition: Web 4 - n. An economic system where users use blockchains to provide content, information, expertise and work to AI agents who then extract value for that work on the users behalf - often by interacting with other AI agents across the world - and pay them for it in cryptocurrency
The term “Agent” is especially fitting here, because you might think of these AI Agents like a Hollywood Agent, negotiating on behalf of creators in an increasingly complex and esoteric technical licensing space.
When you’re “Building in Web4” you’re essentially figuring out how an AI agent can take something that a user creates - whether that’s being in a crypto Telegram groupchat and sharing market intelligence, creating videos or sharing posts (like this one), posting Videos / long form interviews, making money with it - and sharing the pnl with the user.
Humans will continue to be interested in the affairs of other humans, have mimetic desire, and millions of years of programming suggesting we want sweet salty snacks when we’re tired. Our nature won’t change. But that also means that users are not going to abruptly start enjoying the experience of bridging to Arbitrum.
This creates a base case, where - essentially humans ask AI agents to deal with blockchains on their behalf to keep their economic value from being consumed by endless low cost AI agents constantly scraping and copying everything they put on the internet.
This trend will be compounded by the fact that O3 and state of the art models are so powerful, that armed with agentic capabilities, they will create an economy which is essentially unintelligible to human users.
Northwestern University found that over the last 10 years - humans IQ has been dropping for the first time in 100 years in something I call the Flynnstone effect (the reverse of the Flynn effect where IQ used to increase), and this has become more extreme in the past couple years. This mirrors data across the US with rapidly declining test scores and school performance. So
- There will be endless new assets and permutations of AI economic value capture that are increasingly just agents talking to eachother
- This will all be global and impossible to regulate
- It will therefore likely be administered by default on blockchains
- Advanced AI models don’t care how bad the UX is of Web3 products so Web3 is a useful primitive for disputing global IP claims and governance without ever interacting with a government
- Global governments are in fiscal crisis so are cutting their funding anyways making this all more inevitable, especially when you consider there is a technology cold war between the US and China which are both top AI model producers
- The resulting system will be countless protocols, tokens, and computer to computer configuration that will require enormous amounts of intelligence to even understand yet alone use
- People are getting dumber because they’re relying too much on machines to do everything for them
- The largest market cap companies in the world are massive bets on user content creation which will get one shotted by these trends especially when combined with the 5 earlier points about all content profits being cannibalized by video based AI agents
Therefore it is inevitable that
- Users will talk to AI systems
- AI systems will negotiate on their behalf with other AIs
- Users will get paid in crypto to go through the trouble of setting up wallets but the primary interface between users and AI will be chatbots or chat apps like telegram and Discord
Mapping the Current Space From a Web 4 Framework
I think this is a useful category to map the current space.
AI16z and Virtuals for example are Web4 infrastructure. Wheras GOAT is an AI meme coin, and won’t be Web4 unless there’s some sort of payout structure to users interacting with the Truth Terminal. Many TAO Subnets embody Web4 ideas – and TAO itself is an economic system or L1 for Web4 use cases.
There are many AI coins that are simply composability plays - such as Akash and Render, which are currently mostly selling services to human buyers but might adapt their product to license to agents. They certainly plug into Web4 as economic inputs but are not primarily focused on Web4 as such.
Some infrastructure plays would be classified in Web4 - for example, a protocol like Story - which creates a governance process for IP claims on chain. It likely ends up being Web4 because it will become a 2 sided market place between AI agents completely abstracted from any human users.
AI agent browses IP directory, trains a new video on the IP directory pays for it, and then at some point the money (Story tokens, or whatever they get converted into) get airdropped to the user. As you can probably imagine with this workflow, there will be nearly endless token swaps happening at all times to make this intelligible to the end user who doesn’t want a million different currencies - which benefits exchanges like Hyperliquid.
Why Post Fiat Will be the Kingmaker in Web4
The Protocol I’m building - which is called Post Fiat - is solidly in the Web4 space and is an L1. The basic bet here is that you can actually have an agentic scoring process for a large number of AI agents on the network, and use their interactions with users to distribute network rewards using an AI driven process. Eventually Post Fiat’s entire governance process will be run by an AI agent – and thus, rather than the goal being just decentralization - the goal is “Objectivity, Fairness and No Humans Involved”. This differs from Stake based L1s like Solana where ultimately it’s just a bunch of dudes on a Zoom call who claim to be “decentralized” deciding how things are going to be, and where resources are going to flow
Because Post Fiat is designed from the onset to eventually become a vast resource allocation system for AI agents, with superior tokenomics and a clear vision to how it ends up (with AI agents actively securing the network, and an AGI ultimately determining who gets paid) – I believe Post Fiat will end up dominating the Web4 space (btw we are hiring – email me at [email protected] if you’re interested). Other protocols simply have protocol level flaws where big holders of Network Stake will be unwilling to opt into such a system. They also largely assume a human to human world for security assumption rather than what I think is a realistic base case of all network validators in the future being run by advanced AI systems. So it became clear from the onset that Post Fiat could not be a Stake based protocol - which is why we are building on XRP’s core technology (more on that to come).
Zooming Back Out: Everyone Will need a Web4 Plan
Agentic Decentralization will likely become the standard for most protocols (L1s and otherwise) in the Web4 space, and mirrors some proposals currently in progress at Bittensor regarding using AI to reward subnets in proportion to their “scored” value. The question becomes how well these systems work, whether they’ll be adopted by the existing political structures of protocols who probably don’t want something like this to happen, and how appealing it is for AI agents to interact with. Because ultimately that determines how much they can pay users.
Whoever builds the best mousetrap for the AI agents to negotiate value on users’ behalf will maximize the amount of value that users can earn interacting with their agents. This will be the real condition for evaluating L1s going forward. And then – the question regarding AI infrastructure and utility plays is how effective is their interoperability. And the question for exchanges is how seamlessly they can abstract token conversion away from users – who don’t want payouts in 90 different blockchains.
Users want fewer clicks and want to earn a lot of rewards with as little thought as possible. AI is a global movement, and governments simply don’t have the resources to deal with the complexity bomb that’s coming in the next 6 months due to the internet of agents. That’s the core macro behind Web 4. I believe users will be willing to give up substantial upside to AI agents - not because they particularly want to, or think it’s cool. But because it will become increasingly clear this is the only way to make it in the new economy.
Web3 was a cool ideology but Web4 will be a survival mechanism.