Ouija Compiler: Terminal app for dead programming languages
Discover Ouija Compiler, a local AI terminal app running on open-weight models that chats with legacy languages like COBOL and Fortran.
Stock photo for illustration only, not from the actual event
- Ouija Compiler is a terminal app designed for the Hacktoberfest AI Challenge.
- It lets developers converse with legacy languages like COBOL, Fortran, and Pascal.
- The app refuses to answer queries about living modern programming languages.
- Every code snippet is verified through real local compilation and execution before display.
The Hacktoberfest Open-Source AI Challenge Week 4 under the Ghost in the Machine theme has introduced an intriguing submission called Ouija Compiler. This terminal application allows developers to converse with the ghosts of dead programming languages. Users can ask how to implement a loop in COBOL, read a file in Fortran, or build a record in Pascal, receiving responses in spooky prose alongside a complete working program.
When questioned about living programming languages, the veil remains closed as the system refuses to answer while staying completely in character. The creator aimed to honor past technologies that are no longer taught in schools but whose legacy code continues to exist across older systems in production today.
Stock photo for illustration only, not from the actual event
Because small open-weight models can struggle with obscure syntax, reliability is a core feature of the application. Every generated answer undergoes rigorous verification: the code is compiled and executed using a real compiler before the user ever sees it. For instance, asking how to loop five times in COBOL triggers vector searches through scrolls, atmospheric terminal text, and a clean run output showing iterations from 00 to 04.
The Ouija Compiler project brilliantly demonstrates how to mitigate small language model hallucinations by integrating automated testing loops and RAG over verified reference scrolls. By running entirely offline via Ollama on local hardware, it showcases both privacy and creativity, turning mundane debugging into an interactive séance with computing history.
Everything runs locally on the user's machine utilizing open-weight models, ensuring complete offline capability. Developers interested in legacy systems can explore the repository to examine how local AI tools can blend technical utility with creative storytelling.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment