Apprentice: A Slim, Extensible Coding Harness (Written in Common Lisp) (github.com)

Apprentice, a new coding harness built in Common Lisp, has been introduced to enhance the configurability of AI/ML agent systems. Unlike many existing frameworks that offer limited customization, Apprentice allows developers to easily modify and expand its functionality through four key abstractions: Models, Tools, Anchors, and Loops. This architecture promotes ultimate flexibility, empowering users to define how models interact with tools and how data is processed, aligning with the needs of diverse AI applications. The use of Common Lisp brings significant advantages, including macro support for language-native abstractions and instant updates through its image-based runtime. Developers can quickly set up and experiment with various models, from "claude-sonnet-5" to "gpt-5.6-terra," while creating new tools and anchors tailored to specific tasks. The project not only encourages customization but also leverages LLMs for its development, suggesting a collaborative approach to advancing AI technologies. By fostering an environment where both novice and experienced developers can innovate, Apprentice has the potential to accelerate advancements in AI/ML frameworks and enhance the creativity and efficiency of software development in this realm.