Скачать или смотреть Claude vs Codex vs Copilot | Same Task, 3 AI Agents, 3 PRs Compared

In this TechRill deep-dive, we assign the same real-world development task to three AI coding agents — Claude, Codex, and GitHub Copilot — and compare the results end-to-end.

Note: The agents used in this demo—Claude, Codex, and Copilot—are available as part of a GitHub Copilot subscription.

🎯 Task:
Build the landing page of the SkySail flight booking app from a Figma design.

But we didn’t stop at UI implementation 👀
We added advanced requirements:

✅ Clean code structure
✅ Documentation generation
✅ Unit test coverage
✅ Pull Request creation
✅ Integrated checks & validations
✅ Final output verification

In this video, we compare:

🔹 The Pull Requests generated by each agent
🔹 Time taken for implementation
🔹 Code quality and structure
🔹 Documentation depth
🔹 Unit test coverage
🔹 Built-in checks and execution flow validations
🔹 Final output accuracy vs Figma design

This is a practical comparison for developers exploring:
• AI coding agents
• Multi-agent workflows
• PR automation
• Dev productivity acceleration

🧩 Timestamps
00:00 Introduction
01:37 Copilot Creates Issue/Task
03:08 Assign the Task/Issue to Claude, Codex and Copilot Agents
03:55 Verifying the completed Agent sessions - duration & Integrated checks
07:02 PR Verification
07:59 Application Verification - UI matching with Figma design
12:37 Verification of Project Structure, documentation & Unit tests
16:16 Conclusion



If you’re curious how different AI agents behave under the same constraints — this experiment is for you.

🚀 Subscribe to TechRill for more deep dives on GitHub Copilot, AI agents, and next-gen developer workflows.

#AIAgents #GitHubCopilot #ClaudeAI #OpenAICodex #AICoding #AIComparison #DevProductivity #TechRill


Disclaimer: This channel, videos, and content are created in my personal capacity and are a product of me, Ambily K K! They are NOT officially affiliated with, endorsed by, or representative of GitHub (my employer) in any way. All opinions and views expressed are solely my own.