Our Research
Original DataAIHub research — durable syntheses and dated snapshots across agents, documents, models, and GitHub.
When AI Agents Fail
AgentsAs of · 26 September 2026
A successful final state does not necessarily mean the trajectory was reliable.
Read reportState of Open-Source AI on GitHub 2026
GitHubAs of · 17 September 2026
In a 326-repository eligible cohort, the 10 repositories with the most stars hold 14.0% of cohort star mass, while the 10 with the most recorded 30-day commits hold 65.7% of recorded 30-day commit volume.
Read reportDocument Intelligence 2026: Architectures & Evaluation Gaps
Document AIAs of · 29 August 2026
Document AI spans four distinct jobs—parse, schema extraction, document QA, and OCR; public evaluation is increasingly production-shaped but remains fragmented across extraction, completeness, grounding, robustness, and production economics.
Read reportTop GitHub Repositories in the AI Agents Ecosystem 2026
AgentsAs of · 14 August 2026
Hermes Agent leads 63 GitHub repositories classified as agents — 19% of the tracked AI cohort.
Read reportTop Open-Source & Open-Weight AI Models 2026
ModelsAs of · September 2026
Llama 4 leads on open ecosystem and deployability; open weights are not automatically open source.
Read reportTop 20 AI GitHub Repositories of 2026
GitHubAs of · 5 August 2026
OpenClaw ranks #1 among 326 eligible AI repositories on combined influence and activity.
Read report