Essair
Scalability research, 2026. By Rivian Pratama. A document reviewer using a Socratic, broken persona.
Most review tools give a score and a list of problems. This project asks what happens when the reviewer is a person who questions you.
The reviewer speaks through a Socratic persona that answers with questions. The persona is deliberately broken so the critique seems human, and the follow-up generator returns up to three probing questions per reply. The analysis becomes a schema of approach, target section, and priority questions, and the chat loop keeps pressure on the weakest parts of the essay.
I wanted to see if a machine could learn what good writing is and use that to push a writer further.