THE WORKBENCH / INDEX
Experiments
in the open.
Browse the questions we’re exploring. Each project makes its status, intended approach, and current limits clear.
3 EXPERIMENTS
01Planned
AI Safety · Red Teaming
Prompt Injection Observatory
A planned test space for studying how instructions cross boundaries in tool-using AI systems.
02Planned
Systems · AI Safety
Agent Boundary Map
A planned visualizer for mapping permissions, tools, and trust boundaries in AI workflows.
03Planned
AI Safety · Red Teaming
Model Stress Lab
A planned collection of repeatable probes for exploring model behavior under pressure.