TechFlow news, December 31 — Andrew Kang, co-founder of Mechanism Capital, posted that the integration of AI and cryptography presents a revolutionary opportunity for scientific paper review. The crypto project @yesnoerror is developing an AI model to audit 90 million research papers. It has already reviewed over 1,700 papers, identifying an error rate of approximately 3–4%. The AI can accomplish in weeks what would require 45,000 person-years of manual work, at just 1% of the cost of traditional human auditing—around $30 million compared to $5.4 billion.
Kang noted that the project is building models to evaluate paper quality, generating standardized quality scores for each paper by assessing factors such as methodology, logical rigor, and data integrity. This could help distinguish high-quality research and potentially promote better science through rankings of universities and research institutions. Furthermore, the AI model may revolutionize the peer review process and eventually replace human reviewers altogether. Kang stated he has been supporting the project’s operations behind the scenes, emphasizing that titles or positions are not important.
