Claude与Codex协助调校老款Jeep通过检测
开发者Muhammad Zulali借助Claude进行诊断,并由Codex编写代码,帮助一辆因特定车速下氮氧化物超标而未通过加州尾气检测的1994款Jeep Cherokee。Arduino Nano将车辆老旧的发动机电脑连接到笔记本,可读取数据并控制部分设置;项目所用微控制器为3美元,全部元件成本不超过15美元。Tom’s Hardware称车辆最终通过了排放检测,TechRadar则称AI辅助逆向发现了可能与加州相关的燃油供给编程问题。
该项目展示了Claude与Codex协同进行车辆诊断、代码生成和EEPROM编程,是AI智能体协助处理老旧硬件故障的实际案例。
An AI enthusiast managed to get his old 1994 Jeep Cherokee to pass the California smog tests, all thanks to the help of AI and a little electronics wiring skill. This 32-year-old car had only a primitive engine computer, but after connecting it to a laptop via an Arduino Nano board, many settings could be read and eventually controlled.
The above video documents Muhammad Zulali’s journey from the Jeep failing consecutive smog tests, to managing to hook it up and get readings from the ECU, to tuning and fixing the engine emission issues. All the while, Claude provided guidance, and Codex wrote any required code.
This rather old Jeep was failing the California smog tests due to producing too much NOx at certain test speeds. It initially failed at 25mph. Then, after Zulali made some adjustments to the oxygen mix, it failed at both 25mph and 50mph. Yes, it got worse.
Zulali admits he was effectively making adjustments blind, and subsequent regulatory pollution tests cost nearly $100 each time. That’s clearly a big problem, but we have AI to help us in 2026. Actually, it was AI and a $3 microcontroller, contributing to a total bill of materials of no more than $15.