The limits of web3 - The Intersubjective Consensus Problem

I watched this talk right before learning about the transition from Web3 to Web4. Thanks to it, I could understand the limitations of Web3 and the ways people are trying to work around them.

Here is a summary done using NotebookLM. It is good enough to get an idea of the core argument and avoid watching the whole podcast if time is short.

I’m posting this because it’s probably the easiest entry point for understanding the impact of Society Protocol on the Web3 ecosystem. By modeling “energy” instead of “objects” in the blockchain, it opens up a completely new design space for the whole ecosystem.

**
The Core Mission: A “Commitment Superhighway”**

To understand the intersubjective consensus problem, one must first view blockchains not just as “trust machines,” but as commitment engines.

  • Coordination = Communication + Commitments: While the internet serves as a communication superhighway, global coordination requires a way to make and automatically enforce commitments.
  • Verifiable Commitments: Blockchains like Bitcoin and Ethereum began this by enforcing specific rules (e.g., “there are only 21 million Bitcoins”) through code rather than fragile human systems.

The “Truth Gap” and the Topography of Truth

The “intersubjective consensus problem” arises from a gap between what blockchains can see and what humans know to be true. The source identifies three levels of truth:

  1. Mathematical Truths: These are objective and provable on-chain (e.g., “does this private key match this public key?”).
  2. Subjective Taste: These are private facts (e.g., “I like vanilla ice cream”) that cannot be used for broad coordination.
  3. Social/Intersubjective Truths: This is the middle ground. These are facts that “reasonable humans” agree upon, such as who is currently the president or what the price of a stock is.

The Problem: Current blockchains are “blind” to social truths because they only have trustless access to mathematical truths. If a smart contract needs to know a real-world fact (like an election result) to execute, it faces a vulnerability.

The Solution: Intersubjective Consensus

The source argues that even the “objective” Ethereum blockchain actually relies on intersubjective consensus at its base.

  • The Forking Model: If a majority of Ethereum validators tried to lie about a block (e.g., giving themselves millions of ETH), the human community would “fork” the chain. They would collectively decide which chain is “true,” effectively making the malicious actors’ money worthless on the legitimate chain.
  • Generalizing the Solution: EigenLayer aims to use this same principle for other social truths. By using a forkable token, a protocol can adjudicate social truths (like an AI model’s output or a price feed). If the majority of token holders lie about a fact, the community forks to a new version of the token where the liars are excluded.

Real-World Impact: “Jailbreaking” Trust

This technology allows for the creation of Autonomous Verifiable Services (AVSs) that can “jailbreak” data from authoritarian networks.

  • Example: A driver for a major ride-sharing app (like Uber) can use Zero-Knowledge technology to prove their 4.9 rating to a new, decentralized competitor without needing Uber’s permission. This moves “social capital” out of the hands of centralized “Zars” and back to the individual.

The Future: Sovereign AI Agents

The ultimate expression of this is the creation of sovereign AI agents.

  • Currently, AI operates on “delegated authority” from humans or corporations.
  • By using intersubjective consensus, we can create AI agents that own property and enter into contracts independently.
  • These agents would be bound by “constitutions” or coordination conditions that are verifiable and enforceable on-chain—if an agent violates its rules, its property can be “zeroed out”.

In 10 years, the source envisions a world where “swarms” of these autonomous, verifiable agents hold trillions of dollars in value, operating under self-enforcing coordination systems rather than centralized control.

Yeah, this was the part that made it click for me.

EigenLayer is interesting less as a solution than as a symptom of Web3 reaching its limit. It recognizes that some commitments depend on social truth, but because Web3’s native tools are tokens, stake, and slashing, the only available response is to attach capital to social judgment.

That makes sense inside the token ontology. But it also shows why the ontology is inadequate.

A forkable token tries to make social truth costly by financially punishing the wrong answer. Yet social truth is not produced by token weight. Token stake can be one costly signal, but it cannot be the whole social layer. Otherwise, whoever has the most capital gets the most influence over what the system recognizes as true.

That may work for block production. It becomes strange when applied to legitimacy, contribution, identity, or social reality.

The deeper shift is this:

Web3 treats social truth as an input.
Society Protocol treats social truth as part of the state machine.

SP does not magically solve external truth. A blockchain still cannot directly know who won an election or whether a real-world object is authentic. The difference is that SP does not reduce social recognition to token exposure.

Instead of asking, “Which token holders should be punished for reporting the wrong fact?” SP asks:

Which Actors, with which histories, under which Value Functions and governance process, recognized this Event—and what consequences should follow inside the shared state?

That is a completely different architecture.

This is also why the oracle problem now looks different to me. It is not just a missing data feed. It is what the missing social layer looks like from inside the token paradigm.

Web3 built consensus for blocks and mistook it for consensus for society.

Society Protocol is here to fix that mistake.