Skip to content
Agentic MEV Day 2026

NDAI & Information Flow Control in DeFi

Quintus Kilbourn

Session summary

The presenter outlines NDAI, a theoretical mechanism from an academic paper for negotiating with private information without exposing it. They illustrate the problem with OTC trades and potential mergers, where revealing sensitive details to value a deal can destroy its advantage or leak core intellectual property if no agreement is reached. In NDAI, each party encodes its private information into an agent. These agents negotiate inside a Trusted Execution Environment (TEE) that is structured so only a single bit of output (“deal or no deal”) is released, after which the TEE’s memory is wiped to limit information leakage.

The talk then connects this idea to MEV and DeFi. The presenter explains how, as a single block builder applies financial transactions, new arbitrage opportunities appear before transactions are finalized and public, making them hard to capture. They describe an early, small-scale language for submitting many conditional arbitrage transactions, and a proposed TEE-based design where unexecuted, non-public transactions and an arbitrageur’s private algorithm are combined inside a TEE. According to the presenter, this restricts the use of private order-flow data to arbitrage only, with performance illustrations based on non-real, example data.

Key takeaways

  • NDAI is presented as a way for parties to negotiate using private information encoded into agents that interact inside a TEE, which outputs only a single bit (“deal” or “no deal”) before wiping its memory.
  • The presenter argues that in settings like OTC trades and mergers, making proposals public can undermine their value or expose key intellectual property if the deal fails.
  • In a DeFi and MEV context, the authors propose conditional arbitrage transactions and a TEE-based architecture so that private, unexecuted transactions can inform arbitrage algorithms without broadly revealing sensitive data.