How it works
The GitHub archive functions as a sensor for implementation-level and operational friction in software projects. Public issues and pull-request activity are collected deterministically and used to identify recurring technical constraints, deployment problems and compute-related implementation patterns.
The archive is not intended to measure software adoption or repository popularity. It provides a complementary signal to the scientific evidence in arXiv and the attention measurements from Wikipedia.
The archive is accompanied by a complete local PDF collection for the full dataset. The present b>
LLM-assisted analysis
Observatory reports 001–004 showed that many useful questions require examining relationships between signals and their technical context. This led to the development of an LLM-assisted analytical interface over the active Observatory archives.
The interface is used to investigate patterns, compare evidence across arXiv, GitHub and Wikipedia, and answer questions about compute infrastructure and operational friction. It is not a document retrieval service: source material is used internally as evidence for analysis and synthesis, while original source identifiers and links are provided for verification.