Summary
In January, we were able to accelerate out of 2024 and into 2025 with rapid progress across our deliverables. We developed plans for bringing more AI to The Graph and its users. We developed an AI model to automatically document subgraphs to help The Graph’s UX. We pushed the boundaries of cryptography, releasing a crate and white paper based on the work we did to create GraphTally (formerly TAP). We also went further than we have ever gone along the path of making The Graph’s data verifiable, focusing not just on raw data with verifiable extraction, but on data that has been through a full ELTQ pipeline. In addition to all of this, we’ve nearly completed all of the work required for GraphTally to support Horizon. It has been a productive start to the year, and we only plan to grow the team and deliver more from here.
AI
Supporting Crypto AI Agents
Semiotic has been working closely with The Foundation over the past month to strategize how The Graph could better serve AI agents with data. As part of this plan, over the next month, you will begin to see new Substreams emerge that are tailored to common needs of crypto AI agents. You will also begin to see blog posts and tutorials come out around building AI agents on top of The Graph’s data. Further out in the future, we will make it extremely straightforward to build AI agents on The Graph, releasing frameworks for agents to interact with The Graph’s data, building tooling to spin up apps using natural language alone, and even creating Graph agents that other web3 agents can interact with.
GraphDoc
Over the past month, we continued our work on subgraph documentation generation, focusing on implementing a framework for continues improvement as the service goes live and we begin to collect increasing amounts of data. Additionally, we have been focused on reducing the amount of initial hallucinations in the system as we prepare for release. This upcoming month, we will be moving to release the service for utilization by subgraph developers and are excited to collaborate with the other core developer teams in order to integrate it with existing subgraph developer tooling. We will also publish our work in an upcoming blog post.
Verifiability
Verifiable ELTQ
This past month, we explored how our verifiability tools can be applied to The Graph’s ELTQ stack. Our research has centered on evaluating different proof mechanisms and their compatibility with ELTQ workflows, with an emphasis on balancing efficiency, scalability, and security.
More concretely, we created a diagram describing an optimistic protocol we need to build in order to enable verifiable ELTQ. We then narrowed in on extending our work on verifiable extraction (veemon) to support this use case, identifying interfaces and integration points into The Graph’s ELTQ stack. Moving forward, implementing this will be our focus.
h2s2
Over the past month, we advanced the development of Holographic Homomorphic Signature Scheme (H2S2), a cryptographic library designed to enable efficient verification of aggregated signatures. Built on the NCS1 protocol, H2S2 allows multiple signatures to be combined into a single, verifiable signature while preserving individual message integrity. This work builds on the cryptography work of GraphTally.
Our recent efforts focused on refining the implementation within the ark-works cryptographic ecosystem and extending its trait-based design to support additional signature schemes. We also continued validating core functionalities, including key generation, signature aggregation, and precomputed verification. These improvements are foundational for ensuring scalability and efficiency in verifiable data integrity.
You can read more on our blog post.
Indexer Stack
GraphTally (formerly TAP) and indexer-rs
Following up on last month’s progress regarding testing, we have added several new tests to better simulate the flow of the tap-agent component. This includes an integration test for each layer of the actor system, as well as a dedicated test for the tap-agent itself.
We also developed new documentation to aid contributors in understanding the nuances of the codebase. We are cleaning it up and will soon release this publicly.
We’ve begun and made tremendous progress on implementing GraphTally v2 with several components already added to the repository. We are writing the code so as to not interfere with the operation GraphTally v1. As part of this approach, the implementation will remain dormant within the codebase and will be activated alongside Horizon upon its launch.
While the final plan for Horizon is still evolving and some details remain subject to change, we’ve added the core implementation to GraphTally v2. Any future adjustments should be minor once the final specifications are confirmed. We will also begin to adapt indexer-rs more generally for Horizon.