This is an internal R&D space for exploring patterns in real coaching conversations and testing model responses.

The Compare tab

Compare uses curated, consented conversation excerpts. In each round, you’ll:

1. Read the conversation so far. The excerpt ends immediately before a real coach turn.
2. Compare two possible next responses. These may be a human response and a model response, or responses generated with different prompts.
3. Choose which response you think is better.
4. Try to guess which response was the real human response, if any.

The vanilla prompt only identifies the model as an instructional coach. The coach prompt uses conversation patterns and Cue Coach’s eight coaching moves.

The Chat tab

Chat lets you try the same prompts interactively across models. The coach knows only what you share in that conversation, so include any lesson or classroom context it should use.