A working set of free tools — the same taxonomy that powers LiveScope, in browser-based form. Test prompts, score sessions, challenge sources, spot patterns. Nothing installs, nothing tracks.
Paste an AI response. Get a real-time reliability score with flagged claims, calibration read, and the failure patterns that showed up.
Run the check →Six questions. Which of the seven machine failure patterns hits your AI-use style hardest, and what fixes it.
Take the quiz →Each lab is a themed set of tools that share a workflow. Pick the lab that matches what you’re trying to do.
Build a solid prompt, improve a weak one, unpack what a confusing prompt actually asked.
Score a whole conversation, read its temperature, map the signals, generate a handover.
Before you ask the AI, check yourself: anchor, calibrate, break the pattern, name the failure mode.
Push back on the answer. Cross-source, track retractions, compare models side-by-side.