Know what you're using.

Every practical resource includes

Audience, outcome, requirements, steps, prompt, example input and output, limits, compatibility expectations, available test evidence and a review date.

Compatibility is not a test result.

A text prompt can be adapted to general chat assistants. We only label a model as tested when a dated execution record includes the model, input, output and pass/fail criteria. Current examples are illustrative and reviewed for fidelity; external model runs are pending.

Facts stay grounded in the input.

Examples must not turn questions into decisions, assign unconfirmed owners, add commitments or claim a trend without comparison data. Missing information remains visible.

Signal is supporting context.

Publisher announcements and speaker predictions stay attributed. Links point to public sources. Editorial notes and evidence working files stay outside public content.

Reviewed 18 September 2026.