Has anyone here seriously used AI (e.g., Claude for long-context paper analysis, custom GPTs on arXiv, or specialized agents) to aid in hypothesis generation, research gap identification, or drafting substantive parts of a paper?
What are the biggest pitfalls regarding accuracy, hallucination of citations, or superficial understanding of complex theory? How do you validate the AI's output?
Do you see it as a legitimate accelerator for early-stage research, or more of a productivity tool for mundane tasks? Any success stories linking it to a tangible research outcome?
Looking for honest experiences from academics, industry researchers, or solo discoverers.