Tomasz Kornuta, Ph.D.
VP of Engineering / Head of AI
Ethereum and other blockchains support decentralized, permissionless, programs called smart contracts. Smart contracts are combined with off-chain programs to build Web3 protocols. These protocols often have various types of roles, with each role possessing unique responsibilities, rights, and incentives. While real people are normally responsible for fulfilling the requirements of a protocol role, there are often too many factors to take into account for a person to perform irrationally. Furthermore, it is difficult for a protocol designer to predict what will happen when optimal (and perhaps malicious) actors engage with their system. Semiotic Labs uses reinforcement learning and related AI techniques to eliminate these problems by providing simulation and automation tools to protocol designers and human users. We improve the efficiency and security of people and Web3 protocols
Our AI work is focused on improving The Graph protocol. A recent milestone for that effort was the development of what could be the world’s first reinforcement learning agent to autonomously choose prices and compete in a Web3 market. Simulation is a foundation for advanced results in AI. Most of our current efforts are focused on building a simulation capability that both meets the needs of The Graph’s developers and serves as a platform for training more autonomous agents within The Graph. In addition to our work in The Graph, we organize the Workshop on Incentive Mechanism Validation, which brings researchers and protocol developers together to discuss the application of AI for building and improving Web3 protocols.
VP of Engineering / Head of AI
Head of Research & Co-Founder
Senior Research Scientist
Software Engineering Intern
This article explores the principles and mechanisms behind the many popular Automatic Market Maker designs currently used in production. While the mathematical details of these designs are fascinating in their own right, this article seeks to instead focus on graphical representations and high level concepts, allowing for a more approachable and exhaustive exploration of the space.
Indexers within The Graph Protocol are rewarded via an indexing reward. How can indexers optimise their allocations so as to maximise the reward they receive? In this blog post we formalise the problem in terms of a reward function and use convex optimization to find a solution.
Indexers in The Graph have control over the pricing of the GraphQL queries they serve based on their shape. For this task, The Graph created a domain-specific language called Agora that maps query shapes to prices in GRT. However, manually populating and updating Agora models for each subgraph is a tedious task, and as a consequence most indexers default to a static, flat pricing model.
To help indexers with pricing in the relative resource cost of serving different query shapes, as well as following the query market price, we are developing AutoAgora, an automation tool that automatically creates and updates Agora models.
Be the first one to receive the news.
Signup to our newsletter.