Senior AI engineer Helex published a collection of 38 open-source benchmark tasks on GitHub designed for AI evaluation and data annotation across software engineering, machine learning, debugging, and data processing domains. The central repository, afterquery-data-annotation, aggregates these tasks, while three specialized repos demonstrate concrete technical scenarios: afterquery-reckon-sweep-cost analyzes ML training logs, batch-size experiments, and compute-cost trade-offs; afterquery-ledger-rule-races models database concurrency correctness and race conditions; and afterquery-reckon-error-budgets implements SRE-focused error budget calculations. The author frames these resources as practical references for developing reproducible test environments, evaluating AI-generated solutions, identifying edge cases, and applying rigorous engineering practices during model training and assessment. All repositories are personal projects unaffiliated with official AfterQuery documentation.