Developer Uses Claude and a $3 Arduino to Reverse Engineer a 1994 Jeep
Developer Uses Claude and a $3 Arduino to Reverse Engineer a 1994 Jeep
A developer successfully combined the Claude AI assistant with an inexpensive Arduino microcontroller to reverse engineer the primitive computer system of a 1994 Jeep Cherokee. The project ultimately helped the aging vehicle pass strict California emissions tests by diagnosing engine behavior and injecting custom code.
Faced with a 32-year-old Jeep Cherokee repeatedly failing California smog tests due to excessive nitrogen emissions, developer Muhammad Zulali turned to modern artificial intelligence. By connecting a $3 Arduino Nano microcontroller to bridge an outdated vehicle diagnostic port with a modern laptop running Claude, the system extracted live engine telemetry. Despite early setbacks that temporarily worsened test results, the AI configuration successfully diagnosed a programming quirk, analyzed live data during test drives, and deployed custom memory modifications that allowed the vintage vehicle to clear regulatory emissions standards without professional mechanic intervention.
Bridging Decades of Automotive Technology
Modern artificial intelligence platforms are increasingly capable of interacting with physical hardware across vastly different generations of technology. When developer Muhammad Zulali needed to address high nitrogen emissions on his vintage vehicle, standard diagnostic tools proved inadequate for the primitive computer architecture.
Bridging this generational gap required translating modern serial communication protocols into the specific language understood by older vehicle electronics. Using an affordable microcontroller, the hardware setup bypassed the absence of a standard modern diagnostic interface to establish a reliable data connection between the car and the AI.
Chrysler SCI Bus
The vehicle communicated on an outdated serial bus operating at roughly 7,800 baud that required electrical inversion.
Arduino Nano
A low-cost microcontroller handled the necessary signal inversion to connect legacy automotive data with a modern laptop.
Diagnosing and Tuning the Engine
Initial connection attempts successfully retrieved basic engine parameters like coolant temperature and battery voltage, but solving the emissions failure required deeper telemetry. During test drives at twenty-five and fifty miles per hour, Claude logged engine data and fault codes to uncover performance anomalies.
After navigating early diagnostic hurdles and erroneous assumptions about fuel mixtures, the setup revealed a regional programming quirk in the engine computer. This specific quirk reduced fuel delivery under certain scenarios, which inadvertently drove up nitrogen production and caused repeated smog test failures.
Live Telemetry Logging
Claude logged engine performance data and fault codes during active test drives at twenty-five and fifty miles per hour.
Emissions Resolution
The diagnostic process identified a fuel delivery quirk responsible for producing excessive nitrogen during regulatory tests.
Executing Code on Legacy Hardware
Once a credible diagnosis was established, the project faced the limitation that the vehicle's computer predated reflashable firmware capabilities. To implement the necessary adjustments, the developer utilized Codex to write custom code that could be injected directly into a nearby memory chip.
This division of labor allowed Claude to handle complex diagnostic reasoning while Codex managed the memory and code writing tasks. The resulting modifications successfully altered fuel delivery settings enough to pass regulatory inspections roughly a week after the work began.
Open Source Software and Safety Warnings
To share the methods used during the emissions repair, the developer published a free software repository dubbed xj-scanner under the GPLv3 license. This allows other enthusiasts to examine the codebase and understand how modern language models can interact with legacy diagnostic protocols.
Observers note that while the project highlights impressive reverse engineering capabilities, experimenting with vehicle computers carries inherent risks. Artificial intelligence systems remain susceptible to errors, meaning direct hardware modifications can easily introduce new operational problems if attempted without caution.
Why It Matters
This unusual automotive experiment illustrates the expanding boundary of artificial intelligence applications, stretching from digital software suites into physical legacy hardware. By pairing advanced language models with inexpensive microcontrollers, hobbyists can now diagnose and modify complex mechanical systems that previously required specialized dealership tools. However, the project also serves as a reminder that AI tools are imperfect assistants, and directly altering critical vehicle subsystems involves substantial risks that require careful technical oversight.
Someone used Claude and a $3 Arduino to reverse engineer their 1994 Jeep (techradar.com) · 1994 Jeep Cherokee owner hacks 32-year-old car with Claude AI to pass California smog tests (tomshardware.com)
Reporting reviewed through November 14, 2026.
