Scriba
Scriba is one of the two officers of CORTEX, nexmin's AI engine: the clinical notary. It has two offices, describe and compress. It turns an hour of audio into a structured clinical note (SOAP, GIRP, free format) designed to stick to what was said —every statement is checkable against the transcript— and, once validated, it compresses the session into a summary and a title. Describe, don't diagnose, from the paradigm the therapist chooses: that is the rule.
Scriba listens to the session and produces the record. Its first office, describing, is deliberately bounded: it writes the descriptive synthesis of what was said, organises it in the clinical format the therapist has chosen, and pulls metrics straight from natural language (frequencies, repetitions, presence of affect). It sticks to what is in the audio and does not fill gaps with guesses; formulating hypotheses about the case is the other officer's job — Pensa's. That boundary is why CORTEX separates offices. A generalist model that writes the note while also opining on the case contaminates the data with premature hypotheses. If those hypotheses later turn out to be wrong, the clinician can no longer separate what the client said from what the model believed it heard. Scriba protects that boundary by describing, not interpreting. Scriba's second office is compressing: once the therapist validates the note, it reduces the session to a summary and a title faithful to what was approved. Between describing and compressing, Scriba always hands over a draft — in the therapist's format and style — leaving room to edit before closing the record.
Inside nexmin
Scriba is invoked when a session is closed in nexmin (in-person, telehealth or by uploading audio). The draft lands in the clinician's inbox; they edit and approve it before it enters the chart — the Trust Loop pattern. The validated note is the source over which Pensa later formulates hypotheses and reads the process.
Related terms
Last updated: 2026-07-22