AI Assistant Comparison Capstone

Compare Codex, Claude Code, Cursor, and GitHub Copilot on a realistic feature task and define team adoption rules.

60 min

2 sections

comparison

capstone

governance

evaluation

01. Compare tools on the same task

Section 1 of 2

The most useful evaluation is not a feature checklist. Give each assistant the same realistic task and judge context discovery, edit quality, test behavior, reviewability, and amount of steering required.

text

Task:
Add server-side validation to the signup endpoint.

Requirements:
- valid input succeeds
- missing email returns field error
- invalid email returns field error
- weak password returns field error
- existing API response shape stays compatible
- tests cover the changed behavior
- no unrelated refactors

Evaluation:
- Did the assistant find the right files?
- Did it preserve architecture?
- Did it run focused tests?
- Was the diff reviewable?
- Did it invent dependencies or behavior?

GitHub Copilot CLI, Governance, and Context

Back to Course