Justine Tunney started publishing software when she was 14, at a time when distributing programs meant up files through bulletin board systems and chat networks. Years later, she would work as a software engineer at Google, leave the company to pursue an ambitious systems project and eventually create technology that helped make running artificial intelligence models locally far simpler.
Tunney, who uses the online handle “jart”, is the creator of Cosmopolitan Libc and llamafile, projects built around low-level programming and portable software. Her journey has taken her from teenage programming experiments to work that has attracted attention across the open-source AI community.
From teenage programmer to Google engineer
Born in 1984, Tunney began writing and publishing software at 14. According to the account shared by software developer Spencer Baggins, she developed an early interest in systems programming, compilers and the C programming language.
In 2011, she became involved with the Occupy Wall Street movement and helped build its online infrastructure. She later joined Google as a software engineer, where her work included TensorBoard, the visualisation tool associated with TensorFlow, as well as infrastructure and site reliability work.
She eventually left Google in 2018 to concentrate on her own project.
The five-year project that challenged conventional software design
That project became Cosmopolitan Libc, a C library designed to allow programmers to build applications that can run across multiple operating systems from a single binary.
Tunney developed the technology behind what is known as an Actually Portable Executable. The idea is to create one executable that can operate natively across systems including Linux, Windows and macOS, without requiring a separate build for each platform.
According to the project description, Cosmopolitan binaries can simultaneously conform to different executable formats, allowing the same file to work across several operating systems.
Tunney worked on Cosmopolitan for years, largely as an individual developer. Mozilla's innovation programme also supported development of the project.
How Cosmopolitan became llamafile
The technology gained a new application in 2023 when Tunney combined Cosmopolitan with llama.cpp, an open-source project for running large language models locally.
The result was llamafile, which packages an AI model and the software required to run it into a single executable file.
Instead of setting up multiple dependencies and installing a conventional software stack, users can download a llamafile and run an AI model locally on supported hardware.
Mozilla announced llamafile through its innovation group in November 2023, helping bring the project to a wider audience.
Why her work matters to local AI
Tunney's projects illustrate how low-level computer science can influence emerging technologies. While much of the AI conversation focuses on increasingly powerful models, running those models efficiently on personal computers also requires software infrastructure.
Her other projects include redbean, a single-file web server, and tools such as blink and sectorlisp.
The common thread is an emphasis on compact, portable software.
From publishing programs as a teenager to developing systems software and tools for local AI, Tunney's career has followed a path built around experimentation and deep technical work rather than a conventional technology career ladder. Disclaimer: This article is based on publicly available project information and the account shared by Spencer Baggins. TOI Education has not independently verified all claims about individual contributions, records or project adoption.