Everything in the database and every build recipe, unfiltered. The diagnostic narrows this to three; this page does not narrow it at all.
Tools
Loading the database…
Recipes
Things you build or set up yourself. Ordered by ease and safety rather than by fit, since there are no answers to fit them to.
Most of these produce code. A model writes code that looks confident and is sometimes wrong — read the prompt before you paste it, read what comes back, and test the uninstall before relying on the tool.
Loading the recipes…
The rubric
Every filter, weight, penalty and threshold the matcher runs, plus the coverage table including the cells where the answer is that nothing fits. It is rendered from the same file the code reads at runtime, so the published rubric and the running behavior cannot drift apart.
Method and limits
What the research rests on, what it does not support, and the AI disclosure. No user studies, no independent testing, no measurement: every position on this site is argued from the reading and the interviews rather than observed, and the page says so in those words.
Security Rules
Embedded verbatim in every prompt this site generates, and stated as overriding anything asked for afterwards. Rule one is the load-bearing one: It forecloses exfiltration structurally rather than promising against it.
Loading the rules…
Research Context
The synthesis a generated prompt carries with it. Findings only — no interview text, no paper text, and no interviewee named anywhere.
Loading the context pack…
