Start here
A practical workflow for working with AI
CurioLab helps you structure prompts, inspect AI-generated answers, and compare outputs using explicit criteria. Start with the task you need to complete, then use the tools below to make your process more consistent.
Follow this workflow
Write a clearer prompt
Define the task, audience, context, constraints, and output format. Ask the model to label uncertainty and avoid inventing missing details.
Review one answer
Score accuracy, completeness, source fidelity, clarity, and instruction-following. Record concrete evidence for important ratings.
Compare multiple answers
Use the same task and reference material for each response. Score with the same rubric, then export the record for later review.
Read the verification guide
Learn a repeatable way to check claims, find omissions, and separate evidence from assumptions.
What CurioLab's tools do—and don't do
The tools organize your own review and calculate results from the values you enter. They do not independently verify factual claims, certify an AI system, or guarantee that an answer is safe or correct. A higher score means an output scored better under your selected rubric; it is not proof of truth.
For fair comparisons
- Use the same prompt, source material, and requirements for each model.
- Keep evidence notes and disclose how you chose criteria and weights.
- Record model versions and run dates when known; mark unknown details as unknown.
- Do not claim that a small informal test proves one AI is universally best.
Transparency and site information
Before relying on any result, read the site's Editorial Policy, Disclosure, and Privacy pages. The Experiment #001 page describes its own scope and limitations.