Quantum Markets

Animation by Dave Whyte

Background

Today's decision markets can only evaluate a single proposal for each decision. As a result, every new proposal for a decision requires fresh liquidity from traders, leading to a significant capital efficiency problem.

For example, take the following decision: “Which EIP should Ethereum implement next?”

There are 700+ active EIPs that could be considered. Let's say you're a trader with $1M of capital and you have an opinion on all of these proposals. In the standard approach for decision markets, you would need to spread your $1M across all 700+ proposals. This would give you on average less than $1,500 per market to trade.

With quantum markets, you would be able to trade the full amount on each proposal.

Core Mechanism

Let's step through an example of a quantum market that has the goal of finding the best EIP to implement to increase the price of ETH. To those who have heard of–or used–MetaDAO this may seem familiar. In both systems, we take a DAO token (ETH in this case) and allow it to be traded in parallel across different worlds.

However, in the quantum market case we start with a question (”Which EIP should ETH implement next?”) and allow anyone to permissionlessly create (tradable) proposals. With MetaDAO today, we would need to bootstrap new liquidity for each proposal we consider for this decision.

In a quantum market, traders would be able to deposit funds into the system and get an equivalent amount of tradable credits on every current and future proposal for the decision. As they trade these markets, a predicted ETH price would emerge for each proposal.

As time goes on, the market for EIP 2 predicts much higher growth in value for ETH than EIP 1 or 3. When the settlement time is reached, the quantum market triggers a "wave function collapse" by observing the predicted values and selecting the proposal that predicts the highest ETH price.

Since traders were issued their full deposit amount in each market, if a user did not trade in the passing proposal they retain their principal and their PnL is identical to if they had simply held ETH and USDC outside of QM.

Step-by-Step Breakdown

Below we break down a much more granular walkthrough of the EIP example.

Quantum market created with decision criteria:

Select the EIP that maximizes ETH price

The following three proposal markets are created by market participants:

EIP 1
EIP 2
EIP 3

These markets get traded on until they predict an ETH price of $3200, $3000, and $100 respectively. Alice opens the markets and notices that the ETH-EIP-2 and ETH-EIP-3 are both underpriced.

Status Quo Path:

Alice, starting with $1M (500k USDC, $500k in ETH) in the bank, starts depositing funds into the EIP-2 market until the price is $3500. However, the market is liquid enough for these trades to consume all of Alice’s $1M.
Alice does not have enough funds to correct the inefficiency in the proposal 3 market.

Quantum Market Path:

Alice deposits $1M and gets $1M in trading credits in both proposal markets that she’s interested in trading.
She uses her $1M in the EIP 2 market to bring it to $3500, and her $1M in the EIP 3 market to bring it to $3000 (or as close as her $1M will get her).
She leaves the EIP 1 untouched, as she believes it is fairly priced.

End State:

EIP 2 passes because it predicts a higher ETH price ($3500 > $3200).
Proposal markets 1 & 3 are fully aborted/reverted since they didn’t pass. Any trade that happened in those proposal markets is essentially a no-op.
The proposal market for EIP 2 settles an hour after it goes live. The price is $3500.
Alice makes money on this trade since she bought up from $3000 to $3500.

Example 2: Token Launchpad

Today's token launchpads lack any form of accountability. This is especially true in memecoin focused ones that have a notion of "graduating" tokens once they reach a specific market cap. The problem with this approach is that it incentivizes blind sniping, which ultimately muddies the waters.

A second problem is that the space of possible tokens to launch is too large. When you have 100,000+ tokens launching per day, it becomes prohibitively capital-expensive to trade on all of them.

With quantum markets a launchpad can set some cadence with which token launches happen (e.g. one per hour) and let anyone propose an arbitrary number of tokens for each launch. For each new proposed token, all existing traders can trade on its expected outcome (i.e. “Will this token reach $50m market cap within a week?”) without putting up any new capital.

It's difficult to overstate how scalable such a system can get. A QM-based token launchpad can plausibly evaluate millions of tokens for each launch without adding any additional capital overhead to traders. If launch throughput becomes an issue, these markets can be run in parallel batches to accommodate more actual launches.

Proposal Market Flexibility

One important property of quantum markets is that the underlying virtual proposal markets can have any construction as long as markets are directly comparable to each other on some predicted metric. They can be binary or continuous prediction markets, AMM-based, spot, or even distribution market-based.

For example, the EIP example above could have been based on continuous prediction markets for the impressions of a potential tweet. The one that would ultimately pass in this case would be the tweet that has the highest percentage likelihood of YES.

Static vs. Dynamic Decision Markets

Status quo decision markets are static by construction. It is not possible for anyone to contribute a proposal to an ongoing decision market – the only option is to create an entirely new one.

Quantum markets, on the other hand, can be dynamic. Anyone can propose an improved idea into the market while it is active. This is critical for two reasons:

It allows for increased dynamic participation by humans
It allows for unbounded participation by AI agents

The second point is discussed in greater detail in a later section, but it is an incredibly important unlock. AI agents can be much more generative than current decision markets can accommodate, and quantum markets unlocks them to fully express their preferences on-chain.

Since this is a more involved discussion, it is unpacked at the end of this post to not distract from the core mechanism.

Use Cases

The reality is that most important decisions require (or would at least benefit from) an efficient way to evaluate multiple proposals. Many candidate options need to be considered. These use cases are sidelined from using decision markets due to classical markets providing insufficient scale: it is simply too expensive to require fresh liquidity on the long tail of proposals.

Let’s take a small sampling of use cases to demonstrate just how vast the design space unlocked by quantum markets might be:

What should my agent tweet? (Goal: highest expected like/follower count)
What trade should my on-chain vault make? (Goal: highest expected vault token price (proxy for AUM))
Which EIP should Ethereum roll out next? (Goal: highest long-run token price)

Each of these decisions has to take into consideration a large set of options. With the status quo approach, these decisions are confined to only considering a tiny sliver of the option space.

AI Agents and Quantum Markets

AI agents are becoming increasingly capable of making human-level decisions. As a result, there is growing interest in offloading important decisions to agentic AI systems.

Today, most attempts at doing the above have involved hard-coding a specific AI agent as the decision-maker. This model has a number of issues:

Counterparty risk due to centralization from whoever hosts the agent
Agent is prone to mistakes or manipulation that humans wouldn't make
New frontier agents need to manually be switched to (fork or redeployment)

Decision markets are a natural substrate to resolve the issues above:

The best agents make more of the decisions over time as they accumulate capital
Proposals by humans can still be considered as a guardrail to AI-proposed ones
Frontier agents can automatically participate, even before they are publicly known about or open sourced

The issue with decision markets in this context is that they are impractical. Even frontier agents can create proposals at an inference cost that is rapidly approaching $0.

As a result, participation in on-chain decision making by AI agents is constrained primarily by the capital efficiency of the decision market mechanism that they are interacting with.

Since quantum markets have no marginal cost of liquidity for each new proposal, AI agents can run on them at the cost of inference.

For example, a given decision might receive 100,000 proposals from various AI agents, and have each proposal traded on by the handful of frontier agents that can predict proposal performance.

Code

We’ve included starter repos for quantum markets in both Solidity (more in-depth) and Solana/SVM. Our Solidity implementation takes advantage of Uniswap V4 hooks to efficiently handle core functionality.

Disclaimer: these repos are reference implementations and should not be used in a production environment.

Conclusion

Quantum markets demonstrate that the design space of decision markets is still vastly under-explored. If you're interested in working on this mechanism or have related ideas, please reach out on Twitter/X!