Teach the collaboration. Follow the learning.

A common language across a cohort

Course mode applies the same eight behavior axes across lessons or modules. It is designed around individual progress, not comparisons between students.

• Review cohorts using a group overview without rankings.
• See each participant's progress across periods.
• Choose one to three specific actions for the next session.

Feedback with an instructor in the loop

The plugin prepares a draft, with evidence available to the reviewer. An instructor reads and approves it before the participant's personal report includes it. The person card adjusts the form of address, length and tone; it never adjusts scores.

Follow what happens next

Recommendations stay open when the evidence is partial or missing. Later sessions can show whether a learner tried the action. “Not visible” means there is no basis to judge, not that the learner ignored the feedback.

Respect the learning environment

Explain collection, obtain consent and agree how long logs will be kept. Personal files exclude reviewer-only notes and other participants' data. Share group reports deliberately because one file can contain the whole cohort's approved feedback.

A practical origin, without invented outcomes

RetroPair grew from AI-development training workflows. The site uses anonymized demonstration reports with fictional names, not public student results, testimonials or claims of improved course outcomes.

Training pilots — exploring early access

Course mode works in the local plugin. A managed offering is still being explored. Tell us about your format, agent tools, privacy constraints and the behaviors you want to help learners practice.