Same response. Different emotional state. Completely different coaching. An interactive prototype that adapts what good feedback looks like based on how the learner is actually showing up.
Respond to a real workplace scenario, report how you're feeling, and watch the same response generate completely different coaching across three emotional states.
Answer a workplace scenario in your own words — no right or wrong answer.
Slide to where you actually are right now — distressed, okay, or confident.
Get coaching tuned to all three emotional states simultaneously.
Most feedback systems optimize for accuracy — is the feedback correct? But accurate feedback delivered to someone in distress lands differently than accurate feedback delivered to someone confident. The emotional state of the learner isn't noise. It's signal. This prototype asks: what if the feedback system read that signal and responded to it?
You pick your emotional state manually. A production system would infer it from behavioral signals — response time, revision patterns, session history — without asking.
Distressed / okay / confident is a useful starting framework, not a complete emotional taxonomy. Real learner states are messier and more contextual.
This demo primarily adjusts tone and emphasis. A full system would also modulate content depth, pacing, and what it chooses not to say.
Submit the same scenario three times and watch the system override your reported state. Repeated struggle is a behavioral signal the slider can't capture.
What happens when you apply instructional systems thinking to your own workflow — and automate the entire stack with Claude Code.
explore the experiment ← all experiments