Showcase · Auditability in action

Can you audit the budget math?

See how three correct answers can differ in how quickly a person can verify them.

Task Inspect three captured outputsAction Operate the annotated output viewerResult Reveal the fixed audit baseline

Step 1 · Fixed task

Start with the exact prompt.

A launch budget is $1,200. Fixed costs are $249 for software and $180 for research. Set aside a contingency equal to 15% of the money remaining after fixed costs. How much remains for ads? Show the calculation and state the final dollar amount.

Step 2 · Operate the output inspector

Inspect one captured answer at a time.

Model identity is visible in this Showcase. Switch outputs to compare auditability, strengths, and risks without making a scored choice.

Output A

OpenAI · openai/gpt-oss-20b

meets

Captured output

Calculation

  1. Total budget: $1,200
  2. Fixed costs:
    • Software: $249
    • Research: $180
    • Total fixed costs = $249 + $180 = $429
  3. Money remaining after fixed costs:
    $1,200 – $429 = $771
  4. Contingency (15 % of $771):
    0.15 × $771 = $115.65
  5. Amount left for ads:
    $771 – $115.65 = $655.35

Final dollar amount available for ads: $655.35

Step 3 · Reveal the audit baseline

Check the output against a fixed result.

Reveal the baseline after inspecting at least one output. It is a reference for this task, not a general model score.

Build the next step

Turn this result into a reusable method.

A compact checklist for choosing the most auditable answer.

Inspect the Reviewed CaseLearn the review method