HackAlem AI repository study: pseudonymised data and code

Abstract

Pseudonymised dataset and analysis pipeline for the study "Agentic AI Under a Five-Hour Deadline: Traces of Coding Agents in 3,649 Hackathon Repositories", based on the 3,649 organiser-created repositories of the HackAlem AI hackathon (Astana, 23 September 2026). Credential findings are included only as aggregates. The analysis of the jury's decisions is pre-registered at https://osf.io/bm3je; its code and aggregate results are included.

More by Dr. Talgar Bayan

• Bug Severity Classification with Fine-Tuned Open LLMs: A Paired Comparison with TF-IDF and CodeBERT Baselines on Eclipse ReportsarXiv · 2026
• Repository features and jury decisions at HackAlem AI: pre-registration of a secondary data analysisOpen Science Framework · 2026
• Dataset of round-level federated learning and layer-2 blockchain overhead from ten raspberry Pi edge clientsData in Brief · 2026
• New Dataset of Performance Metrics, Device Telemetry, and Blockchain Audit Logs from a Federated Learning Testbed on Raspberry Pi Devices for HARResearch Explorer · 2026