LLM Agents
Read note ↗
AI Auto-Research and Automated Paper Writing: Systems, Open-Source Frameworks, and Empirical Failure Modes
A ground-truth survey of 50+ autonomous research papers, open-source codebases, and benchmarks (2023–2026), audited against 5,932 controlled runs on Judge Goodhart, parent-lineage memory, harness leakage, multi-hop citogenesis, cache-miss economics, and scope-typed Data-Chaining.