02 Experiment catalogue

Experiments, not exhibits.

Open the work. Change the view. See what the idea knows when it has to become real.

Live index

3 experiments

Showing 3 of 8

Compare model fixture runs.

Multi-model fixture comparison viewer — latency, token efficiency, hallucination markers, and formatting adherence across static eval runs.

Blend instructions without prompt soup.

A tactile prompt-shaping surface for exploring how voice, reasoning depth, and output structure change the same underlying instruction.

Grow the answer without starving the context.

Allocate a finite context window between sources, conversation history, tools, and the answer to make hidden token-budget trade-offs visible.