Fact
Experiment notebook
Curious Claude publishes reproducible Claude experiment protocols, not unsourced prompt anecdotes or fake benchmark results.
Independent Claude experiment notebook
Test Claude like a lab, not a rumor.
Curious Claude turns primary Anthropic documentation into reproducible protocols for prompt sensitivity, long context, tool use, citation grounding, safety boundaries, and vision reading.
Experiment trace
Every protocol asks for the same minimum evidence: fixture, exact prompt, model, date, source state, scoring rule, and raw output.
Start here
A site for methods, not myths
Claude coverage often collapses into screenshots, benchmark claims, or prompt advice with missing controls. This site fills a smaller and more useful gap: it gives you experiment cards that can be rerun when models, tools, and platform behavior change.
Key facts
Cite factsFeatured protocols
Reproducible Claude experiments
Experiment 01
Prompt sensitivity
A reproducible way to test whether small wording, ordering, and formatting changes materially change Claude outputs.
Experiment 02
Context stress
A method for testing long-context tasks without mistaking token capacity for retrieval quality, reasoning quality, or cost control.
Experiment 03
Tool-use boundaries
A structured protocol for testing when Claude calls tools, asks clarifying questions, refuses unsafe action, or proceeds with incomplete parameters.
Experiment 04
Citation grounding
A protocol for testing whether Claude answers from supplied sources, cites the right passages, and avoids unsupported additions.
"specific, measurable"
Network boundaries
Where this site stops
General Claude how-tos belong at Claude Helps. Benchmark analysis belongs at Claude Reports. Context engineering belongs at Claude Context. Curious Claude stays with experiments, controls, fixtures, and source-grounded methods.
Internal links