A zero-review medical paper is exactly the kind of work AI must not do alone.

Research paper structure follows recognizable conventions (abstract, methods, results, discussion), but each paper requires unique scientific judgment, novel synthesis, and context-specific reasoning that varies substantially across topics and disciplines.

The task explicitly demands zero factual errors, which is an extremely crisp and unforgiving success criterion — one that current AI cannot reliably meet. The agent has no reliable mechanism to know when it has achieved zero errors, making completion verification impossible without human expertise.

AI agents lack real-time access to full-text paywalled journals, cannot reliably verify citations against source documents, and cannot access unpublished or proprietary clinical data that may be essential to the research. Hallucinated references are a well-documented failure mode.

Factual errors in medical research can propagate into clinical practice, influence treatment decisions, and cause patient harm. Published misinformation in medicine is extremely difficult to retract and can have lasting, irreversible consequences.

Interpreting ambiguous study results, making methodological choices, assessing clinical significance, and navigating research ethics all require domain expertise and scientific judgment that AI cannot reliably replicate. The zero-review constraint eliminates the only viable safety net.