SI Agent News
DevelopingClaudeOpenAI

AI helps owner pass emissions test on a 1994 Jeep Cherokee

A YouTuber used Claude to decode a vintage vehicle's communication protocol and reflash its computer to fix a fuel-mapping issue.

Papercut shadow box: picture for the story 'AI helps owner pass emissions test on a 1994 Jeep Cherokee'
AI-made papercut shadow box for this story. No real people or products are shown.
$3cost of the Arduino Nano used

The short version

Why it matters

This demonstrates how large language models can act as technical consultants for complex hardware troubleshooting. It shows that AI can bridge the gap between legacy mechanical systems and modern digital repair methods.

The challenge of legacy vehicles

Modern cars are generally easy to diagnose because they use a standardized OBD-II port. This port allows mechanics to plug in a generic scanner to see exactly which sensors are failing or which codes are being triggered by the onboard computer.

However, older vehicles like the 1994 Jeep Cherokee XJ do not have this standard connection. Because of this, the owner, Muhammad Zulali, could not use a standard tool to find out why his car failed California emissions tests. He noted that while a Chrysler-specific tool existed for that year, it was difficult to find and provided very little useful information.

This creates a significant barrier for owners of vintage cars who want to maintain them according to modern standards. Without a clear way to read the data from the car's computer, finding a specific mechanical or software fault becomes a process of trial and error. Zulali decided to use artificial intelligence to find a workaround for this lack of diagnostic infrastructure.

Decoding the car's language

To get information out of the Cherokee, Zulali turned to Claude to figure out how the car's computer actually communicates. The AI was able to identify the specific communication protocol the vehicle used to send out data. This was the first major hurdle in making the car's internal systems talk to a modern device.

Once the protocol was identified, Claude suggested using an Arduino Nano to act as a translator. An Arduino is an open-source microcontroller, which is a small computer chip that makers use to build custom electronics and tools. These devices are popular in the hobbyist community for their flexibility in connecting different types of hardware.

Zulali purchased the Arduino for $3 and used it to begin pulling readings from the Jeep's computer. While the initial data provided more questions than answers, it gave him a baseline of information that he could analyze. This setup allowed him to move from guessing what was wrong to actually seeing the numbers the car was producing in real time.

AI as a diagnostic partner

As Zulali began testing different components, he discovered that Claude could do more than just read the data passively. Based on the readings, the AI suggested that the car might have low fuel pressure. When Zulali asked the model to run the fuel pump, Claude was actually able to crank up the pressure on its own through the connected system.

However, the AI's first diagnosis was incorrect because it was based on a faulty dataset. This highlighted a limitation of using AI for diagnostics: the model can sometimes provide confident answers that are not grounded in the correct facts. Despite the error, the process showed that the AI could interact with the vehicle's hardware to perform physical actions.

Zulali continued to use Claude to take readings while he was actually driving the vehicle. By plugging his laptop into the car and running the AI through the Arduino, he could gather a more comprehensive set of data. He eventually had the AI copy the entire computer code so he could examine the logic outside of the vehicle's environment.

Reflashing the computer

The investigation eventually revealed that the Cherokee was failing its emissions test due to a specific fuel-mapping issue. Fuel mapping is the instructions that tell the engine how much fuel to inject at different speeds and loads. Because the mapping was incorrect, the car was not burning fuel efficiently enough to meet state requirements.

To fix this, Zulali needed to rewrite the code and put it back into the car's computer. This was complicated because the older hardware lacked the ability to be conventionally reprogrammed. He had to use a second AI tool, OpenAI's Codex, to help him reflash the computer with the corrected instructions.

By combining the diagnostic capabilities of Claude with the coding strengths of Codex, Zulali was able to successfully update the car. With the new code in place, the Cherokee finally passed its emissions test. The entire process was completed without the owner ever having to visit a professional repair shop.

The barrier to entry

While this project is a success story for AI, it is important to note that it required a significant amount of technical knowledge. Zulali had to understand how to interface microcontrollers with vehicle computers and how to interpret raw data streams. For the average car owner, the complexity of this setup might be overwhelming compared to simply taking the car to a mechanic.

Most people would find it much easier to pay a professional to suss out the problem using specialized equipment. However, the project serves as a proof of concept for what is possible when AI is used as a technical consultant. It shows that AI can help navigate complex, non-standardized systems that were not designed for modern plug-and-play diagnostics.

Looking forward, this suggests a future where hobbyists can use AI to maintain and repair older technology. As more people gain access to these tools, we may see a rise in DIY repairs for complex machinery. It highlights a shift where AI helps bridge the gap between legacy hardware and modern software capabilities.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by The Drive · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news