AI enthusiast Muhammad Zulali managed to tune his 32-year-old 1994 Jeep Cherokee SUV to pass California's stringent emissions tests using the Claude and Codex AI models and a kit of electronics costing less than $15.
AI enthusiast Muhammad Zulali managed to tune his 32-year-old 1994 Jeep Cherokee SUV to pass California's stringent emissions tests using the Claude and Codex AI models and a kit of electronics costing less than $15.
The old SUV was failing environmental inspections due to elevated levels of nitrogen oxides at certain speeds. Initially, the car failed the test at a speed of 25 mph (about 40 km/h), and after Zulali's manual attempts to adjust the fuel-air mixture, the situation only worsened - the indicators exceeded the norm by both 25 and 50 mph (80 km/h), writes Tom's Hardware.
Since he had to act blindly, and each recheck cost almost $100, the car owner decided to connect artificial intelligence and a microcontroller to the diagnostics for $3 (the total cost of all components was no more than $15).
The main obstacle was the incompatibility of a modern laptop and the primitive electronic control unit of the 1994 car. Following Claude's clues, it turned out that the engine uses the Chrysler SCI bus with a data transfer rate of about 7800 baud and an electrically inverted signal.
To solve this problem, the enthusiast built a circuit on a breadboard with an Arduino Nano microcontroller, connecting it with three wires to the Jeep's 6-pin diagnostic connector. After successfully connecting, the system initially only read coolant temperature and battery voltage, but Zulali later configured it to collect advanced telemetry and error codes during test runs at 25 and 50 mph, and was able to control parameters such as fuel pressure.
When the cause of the malfunction was found, it turned out that the main ECU could not be flashed - the car's computer turned out to be several months older than the models that supported firmware updates. Then Zulali brought in Codex, which helped implement and run the custom code through a nearby EEPROM memory chip.
“Claude did most of the diagnostic work, and Codex did most of the memory work. I just connected the wires and drove the car,” Zulali said, adding that he was able to do without the services of mechanics and expensive Chrysler dealership equipment.
The enthusiast made his developments and diagnostic code publicly available on GitHub under the name xj-scanner under the GPLv3 license.
