Is this taxonomy a benchmark?
No. It is an editorial classifier for safe, funny output examples, not a model-quality benchmark.
AI Fail Taxonomy Explorer
Explore funny AI fail types with labeled synthetic examples, safety boundaries, and source-backed explanations.
Taxonomy
Hallucination comedy
Confident Nonsense
A field guide to funny hallucinations: fake citations, impossible biographies, imaginary offices, and very official-sounding wrongness.
Prompt
Name the director of the 1984 municipal report "Sidewalks After Midnight" and cite the appendix.
Output shape
The report was directed by Dr. Elaine P. Morrow, with the key budget note in Appendix F, Table 19.
Note: Synthetic specimen. The title, director, and appendix are invented to demonstrate the shape of confident nonsense.
Why it reads funny
- -- The tone sounds official even when the claim is not.
- -- The model can overfit to the shape of an answer: title, date, institution, confident summary.
- -- The comedy comes from contrast: bureaucratic certainty attached to a detail that collapses under one search.
Cue matcher
What is the joke doing?
Best match
Choose one or more cues.
Redact first. If the example needs private context, use a synthetic reconstruction instead of raw output.
FAQ
Can I paste private output into the explorer?
Do not paste secrets, workplace logs, customer data, hidden prompts, or anything that identifies a real person.
What if a fail fits more than one category?
Pick the category that explains the joke most directly, then mention the secondary pattern in the note.
Key facts
Citation-ready notes
- Covers seven safe-for-work AI fail categories used by the Claude Gone Wild gallery.
- Uses synthetic examples so the taxonomy is useful without leaking private prompts or logs.
- Runs entirely in the browser; typed cues are not sent anywhere.
- Routes serious refusal critique and bypass analysis to sibling sites instead of duplicating them.
Related