The project began when Zulali's Jeep Cherokee failed California emissions testing due to nitrogen oxide levels. Because the 1994 model lacks an OBD2 port, standard scanners were unable to read the vehicle's data.
From reading data to controlling the vehicle
After consulting Claude Code, Zulali discovered the engine computer used an old SCI protocol with an inverted signal. To bridge the gap, he used an Arduino Nano and three wires. While the first attempt failed due to resistors interfering with the signal, the second attempt successfully retrieved real-time data such as coolant temperature and battery voltage.
The AI's capabilities expanded during a night drive. After noticing the computer reported speed at twice the actual rate, Claude corrected the calculation. The AI even took control of the vehicle's hardware, successfully activating the fuel pump for approximately 10 seconds upon request.
Context
The 1994 Jeep Cherokee is a 32-year-old vehicle that requires specialized software like ChryslerScanner for diagnostics, which the AI agents were able to replicate through custom code.
