MEV tl; dr
MEV refers to the "Maximum Extractable Value", an economic phenomenon representing profit opportunities arising from on-chain transactions.
For example, a large swap on a DEX can cause token prices in a pool to fluctuate. Traders might explore the price difference in another pool to generate a positive surplus, and such arbitrage opportunities are considered MEV. In general, arbitrage is an example of healthy MEV, as finding better quotes is vital in a free market and benefits the end user.
On the other hand, toxic MEV can be detrimental. For example, users may unknowingly trade with high slippage and fall victim to sandwich attacks.
You can find more resources about MEV here.
Case Study: Toxic-MEV on Solana
We have seen an explosion of toxic-MEV on Solana, including sandwichers, backrunners, and snipers (bots that congest the network by abusing spam techniques).
On Solana, block building is a continuous high-speed process with no in-protocol mempool. After Jito deprecated its out-of-protocol mempool in early 2024 (which slowed down transaction processing by 200ms and allowed typical MEV attacks and transaction ordering), attackers now have to be close to validators or RPC nodes (side-dealing) to receive the most accurate state of the network while also observing incoming user transactions. This created new dynamics in the MEV space.
Another approach, MEV-protective RPCs, uses various mechanisms (e.g., routing via private mempools or back-running user transactions to return the MEV to users) to minimize toxic-MEV. These out-of-app services create unwanted external dependencies and require users to switch to appropriate RPCs manually.
Additionally, although initiatives such as the QUIC protocol, Stake QoS, or priority fees could help mitigate the problem, incentives for bots will always remain as long as MEV opportunities exist.
At the application layer, Jupiter, the leading aggregator on Solana, has a few initiatives to help reduce MEV attack vectors for its users:
Splitting orders into smaller sizes and statistically optimizing slippage.
Providing the option for users to automatically add a Jito tip to their transactions, which will route transactions via Jito bundles directly to validators, as opposed to public RPCs that can potentially leak information to searchers.
However, these solutions are not comprehensive, as users have to trust that validators act honestly and are not side-dealing with searchers. Adding Jito tips is an additional cost imposed on users. Although the minimum Jito tip is only $0.00015, assuming $150/SOL, the average tip is around $8.80, which can quickly add up.
Currently, Jito MEV tips are a good way to estimate Solana MEV activity (toxic or non-toxic), which shows that agents are willing to pay extra to validators to prioritize their transactions.
For sandwich attacks, sandwiched.me is a good source.
Urani's Approach
Urani addresses the toxic-MEV minimization issue at the application layer by:
Additionally, we coined the term "MEV agents" to distinguish "good" bots from toxic MEV bots. We refer to the humans behind these bots as "MEV agent operators." At Urani, they are as important as any other actor in the supply chain.
Last updated
