Section 077 · Chapter 10, Anti-Patterns That Create False Confidence
Anti-Patterns: The One-Run Demo Fallacy
A beautiful demo proves what the system can do once, not what it will do reliably.
RAGone-run demodemo fallacy
What to do
- Use fixed prompts, documented model settings, versioned tools, recorded traces, and enough samples to estimate behavior.
- Define runnable checks that exercise RAG, one-run demo, and demo fallacy.
- Set acceptable outcomes and blocker failures for RAG, one-run demo, and demo fallacy before running the evaluation.
Evidence to preserve
- Preserve the inputs, versions, configurations, raw outcomes, and results for RAG, one-run demo, demo fallacy needed to reproduce work on Anti-Patterns: The One-Run Demo Fallacy.
- Report results for RAG, one-run demo, demo fallacy by relevant slice, separate blocker failures from averages, state uncertainty and blind spots, and connect the result to a release decision.
Expert note
When the system matters, separate capability demos, smoke tests, benchmark runs, and release evals. A demo can inspire investment, but only repeated, sampled, versioned evidence should support shipping.
Continue the conversation
Apply this to your context.
Save your product context once, then open a focused conversation that combines it with this concept.
Cite this page
Jason Arbon. "Anti-Patterns: The One-Run Demo Fallacy." Testing AI Knowledge Edition, section 77.
https://jarbon.ai/testing-ai/knowledge/ch077-demo-fallacy.html