In January, we:
- Further simplified the design for Verifiable Payments.
- Released updates to AutoAgora and the Allocation Optimizer.
- Began interviewing candidates for our DevOps Engineer and Backend Developer positions.
Looking back (what was delivered)
We have released several improvements to AutoAgora:
- Added the option to exclude subgraphs from AutoAgora: v0.7.0
- Added automatic container builds (per @indexer_payne’s suggestion), hosted at
- (Breaking) Overhauled the arguments interface and handling: v1.0.0
- Added support for multiple
indexer-serviceendpoints (thanks @ellipfra !): v1.1.0
We have been working hard on the release of AutoAgora v2. This release will migrate our learning backend from Torch to JAX, which is much easier to support. It will also set us up for further experimentation. The release supports multi-product pricing (per-query-type automated pricing) and better experiment tracking via sacred. We have also been updating our documentation and documentation generation on AutoAgora to help indexers better use and understand the code.
In parallel with AutoAgora v2, we’ve been working on v2 of the Allocation Optimiser. This release comes with some major performance improvements. In our (far from conclusive) testing so far, we have found an improvement from 5 minutes to optimize to around 0.1 seconds! Note that that time is just for the optimization part of the software. Querying the data has been, and remains, the slowest part of the problem. Other than performance improvements, we also spent some time focusing on the UX. More on that in a later update.
We have drastically simplified the design needed to achieve verifiable payments. The new design is called the Timeline Aggregation Protocol (TAP) and uses no new cryptographic elements. More details on this protocol will be released in the coming months.
Looking ahead (upcoming priorities)
We are looking forward to releasing v2!
v2 coming soon! We’re also working on a technical publication for the optimizer. Hopefully, we will have some updates on that next month.
Verifiable Indexing & Verifiable Queries
We began investigating some game theoretic approaches to solving the VI/VQ problem. This is still a new topic for us, so we don’t have much to share yet. Just know that we’re on it!
Implement TAP prototype using Rust and Solidity.
Large Language Model research kickoff
Advances in LLM techniques, e.g. ChatGPT, could significantly lower the barrier to entry for developers and users of The Graph. We will start preliminary research on this topic.
We are hiring!
Semiotic is hiring a DevOps Engineer and a Backend Developer. Please share this link with your network: https://semiotic.ai/jobs/.