Claude AI resurrects 1994 Jeep Cherokee with a $15 DIY hack
A garage experiment with surprising results.

Claude AI under the hood of a Jeep Cherokee. Image by Cybernews.
- A 1994 Jeep Cherokee passed California smog tests after its owner used AI to diagnose emissions problems.
- Claude helped decode the Jeep’s old engine computer and analyze readings from a homemade diagnostic setup.
- The owner built a basic signal converter for about $3 and completed the project for $15.
- OpenAI’s Codex helped add custom code when the Jeep’s computer was too old for a normal software update.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
An ageing Jeep with an outdated engine computer, a failed emissions test, and no easy way to diagnose the problem becomes an unlikely AI success story.
A 1994 Jeep Cherokee owner has used Anthropic’s Claude AI and OpenAI’s Codex to help him crack the car engine's computer.
AI engineer Muhammad Zulali built his own system for $15 after his Jeep initially kept failing California's smog test.
The vehicle was producing excessive nitrogen oxides, which is deemed to be both a public health danger, as well as a gradual mechanical fault.
And with each failed test costing $100, repeated trial-and-error visits made for an expensive exercise. What’s more, Zulali didn’t have access to the diagnostic data either, which spurred him into action with AI.
Teaching an old Jeep tricks
Zulali tasked Claude with deciphering how to communicate with the Jeep's ancient engine control unit (ECU).
The main obstacle at first was that a modern laptop couldn't be plugged into the vehicle due to electrical signalling being inverted. That effectively meant the 1994 system was flipping its ones and zeros, which a modern device sees as gibberish.
Claude then intervened by inventing a tiny microcontroller named Arduino Nano, essentially a DIY signal converter.
What consisted of a handful of wires and a basic circuit board totalled around $3 to put-together. Zulali connected the three wires to the six-pin diagnostic port of the Jeep.
At first, the only info that could be communicated between systems old and new was coolant temperature and battery voltage.
As the project gained momentum however, Claude helped analyse the readings and fault codes for why the Jeep was producing excessive pollutants.
Calling in an AI tag-team
Claude generalized its findings, as AI models typically do, by comparing them with results from other engines.
Zulali intervened and told the agent to be more specific. The AI also helped explore adjustments to engine controls, including fuel pressure, as he worked towards a credible diagnosis.
Another stumbling block emerged, as the Cherokee’s computer was a few months too old to allow for a software update.
Within touching distance, Zulali called upon OpenAI’s Codex coding agent, sneaking a custom mini-program onto a neighboring memory chip and getting the car to run it anyway.
Zulali explained the division of labor, giving AI a lot of the credit:
Claude did most of the diagnostic work and Codex did pretty much all of the memory work. I just did the wiring and the driving.
The Cherokee was finally able to pass the smog test, while Zulali posted his project online, so anyone can follow suit, should they wish.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.