Climate scientist Alek Petty has published a compelling argument that the traditional peer review system is failing and must be replaced by a hybrid AI-human model. In a recent blog post, Petty details how the publication process was already struggling before the current AI explosion, with editors struggling to find reviewers and papers getting stuck in typesetting for over a month. He argues that the field is moving too fast to wait for the slow, bureaucratic machinery of traditional journals.

The Pre-AI Bottleneck

Petty notes that even before generative AI flooded the inbox, the system was broken. He highlights that finding editors and assigning papers takes weeks, and the volunteer nature of many non-profit journals means that pushing a paper through the pipeline is often a low priority for editors. This structural inefficiency has led researchers to look for alternatives, inspired by AI frontier labs that brazenly post preprints and move on, prioritizing speed and accessibility over formal validation.

Real-World Alternatives Emerge

Petty points to specific examples where the shift is already happening. He cites the Climate Brink Substack by Andrew Dessler and Zeke Hausfather as a valuable, non-peer-reviewed outlet that provides data-driven insights while maintaining scientific credibility through active community engagement. Within traditional journals, he highlights Earth System Science Data (ESSD), which now offers a 'living data' path for regularly updated datasets. This model reduces redundancy by focusing on changes and linking to prior papers, though it still involves formal peer review.

AI Disclosures Are Box-Ticking

The arrival of AI tools has exacerbated the review burden, creating a deluge of papers that are technically polished but often unoriginal or poorly framed. Petty criticizes current journal requirements for 'AI disclosures,' which he sees as performative and useless. He observes that most disclosures are brief statements like 'Claude used to help with data processing,' which fail to explain how AI actually accelerated the science. He argues that if researchers are using powerful tools, they should provide clear guides on how those tools were leveraged, rather than just checking a box.

The Human Cost of Automation

Beyond workflow efficiency, Petty shares a personal reflection on the changing nature of scientific writing. While he enjoys the assistance AI provides across his research workflows, he admits to feeling a lingering sadness that his previously strong ability to draft high-quality scientific papers has become 'nearly meaningless.' This human interest angle underscores the emotional and professional stakes of the transition, suggesting that while AI can handle formatting and initial drafts, the loss of drafting skill represents a cultural shift in how researchers view their own craft.

Predicting a Hybrid Future

Petty predicts that the current peer review system will not survive and will be replaced by a hybrid system where AI handles reproducibility testing and claim vetting, while humans focus on high-level evaluation. He suggests that the prestige associated with fancy journals like Nature and Science will die a death, replaced by new publishing models driven by AI-savvy editors. These new models would prioritize efficient summaries generated by AI, allowing humans to focus on the critical thinking aspects of research dissemination.

Key Takeaways

  • Traditional peer review is already broken, with significant delays in editor assignment and typesetting.
  • Current AI disclosures in journals are minimal and fail to provide useful context on tool usage.
  • Specific alternatives like the Climate Brink Substack and ESSD's 'living data' path demonstrate viable models for faster, more transparent sharing.
  • Researchers are experiencing a sense of loss regarding traditional drafting skills as AI tools become ubiquitous.

The Bottom Line

The academic publishing complex is bloated and slow; AI is forcing a necessary, if painful, evolution toward speed and transparency. If you are building developer tools for science, focus on reproducibility and clear AI usage documentation, not just manuscript formatting.