Recursive analysis through procedural semantic programming
- May 22
- 21 min read
Like a student rushing a single draft of a paper before it is due, generative AI models write once and neglect to check their work. By creating a system prompt to instruct the model how to write, review, and revise on its own, we can mitigate error propagation and improve the caliber of responses. In the methodology for recursive analysis though procedural semantic programming and its template, the Datum Research Analytical Process (DRAP) v1.0, we use natural language to teach models how to (figuratively) think before they speak.