GPT-5.3 Codex vs Claude Opus 4.6 – Coding a Physical Skateboard Game

Описание: Timestamps:

00:00 - Intro
00:50 - Test Overview
01:46 - Prompt Overview
05:21 - Plan Mode Begins
06:19 - Plan Mode Questions
07:47 - Claude Web Searching
08:19 - Codex Implementation Plan
09:13 - Claude Implementation Plan
09:31 - Coding in Process
10:58 - Arduino Troubleshooting
11:35 - Bluetooth Device Setup
11:59 - Codex C++ Build Assistance
12:25 - Codex Skateboard Game V1
14:25 - Codex Skateboard Feedback
14:38 - Claude Arduino Script Issues
16:12 - Claude Skateboard Game V1
18:22 - Game Improvement Prompts
19:44 - Claude Skateboard Game V2
21:50 - Codex Skateboard Game V2
22:45 - V3 Game Feedback & Coding
23:27 - Codex Skateboard Game V3
23:50 - Claude Result Improvement Help
24:52 - Claude Skateboard Game V3
25:39 - Codex Skateboard Game V4
26:27 - Codex Telemetry Improvement Help
27:18 - Usage Metric Information
28:03 - Physical Result Testing
28:11 - Codex Game Result Playing
28:40 - Claude Game Result Playing
29:16 - Closing Thoughts

AI Integration & Consulting https://bijanbowen.com
Join The Discord:   / discord  

In this video, we put GPT-5.3 Codex and Claude Opus 4.6 into a real-world agentic coding showdown—with hardware involved. Using the Codex CLI from OpenAI and the Claude Code CLI from Anthropic, we task both models with building a skateboarding game controlled by a physical skateboard.

The skateboard acts as a real controller via an Arduino with a gyroscope, sending motion data to the game. Each model must plan the system, write code, troubleshoot hardware and Bluetooth issues, iterate on gameplay logic, and improve results across multiple versions.