MASSACHUSETTS — U.S. Air Force cadet Joshua Lynch developed an AI-powered application, ROMAD-AI, during a three-month project with the U.S. Department of the Air Force–MIT AI Accelerator's Phantom Program in Massachusetts. The project involved a process called "vibe-coding," where Lynch guided generative AI chatbots to write and refine code.

Lynch, who had no previous coding experience, utilized paid models of three AI chatbots—Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini—for code generation. Most of the coding work was conducted through the chatbots' main chat functions on a web browser. He also completed several professional development courses in AI during the project.

Lynch initially sought to create an application that would reduce collateral damage and enhance survivability for his tactical team. However, due to limitations in AI capabilities and development time, he re-scoped the project to focus on basic document processing. The final application, produced using Google AI Studio App, analyzes tactical maps and generates mission-planning documents through an interface with a VLM-powered chatbot.

Laura Niss, a technical staff member in the Embedded and AI Systems Group at MIT Lincoln Laboratory, mentored Lynch during the project. "The Phantom student wanted to see if he could create a useful application through self-identified vibe-coding, without any previous experience," Niss said. "Within this project, I wanted to understand how his perception of AI changed over time with use."

Lynch observed that Claude demonstrated more stability across traits such as likeability, anthropomorphism, and perceived intelligence compared to ChatGPT. He also found that the final prototype was not secure for its intended use, as it sent input documents to a Gemini AI model for analysis instead of processing them locally on his computer.

"I was quite impressed with this final product, and it showed me how powerful these systems can be at prototyping designs from nonexperts," Niss said. "I'm now of the opinion that these can be powerful tools for nontechnical experts to convey problems and possible solutions to technical experts, and aid in communicating desired outcomes."