Recommendations need evidence

Keep “looks useful” separate from “independently tested”.

Current preview status

None of the current entries has completed our independent testing. Red and blue badges demonstrate the visual design; they are not actual ratings.

What each test records

• Template source, version, test date, and conditions.
• Input material, full prompt, actual output, and screenshots.
• Elapsed time, observable usage, human intervention, and failures.
• Suitable tasks and limits. Unmeasurable data remains unknown.

How tiers will work

Top pick: strong results in the tested tasks and samples, supported by evidence. Recommended: useful under stated conditions, with human review. Not tested: insufficient evidence. Every color includes a text label; mascot colors do not imply a rating.