Vivien Henz claims that Common Lisp is now the best programming language because its live image system eliminates the traditional compile‑run feedback loop. In most languages a change requires recompiling and restarting the program, which can take minutes, but in Common Lisp a new function definition replaces the old one instantly in memory, so LLMs can see the effect of their edits immediately.
The language’s condition system further speeds experimentation: instead of crashing on an error, Common Lisp enters a debugger that preserves the full stack and all variables, allowing an LLM to inspect the failure, apply a fix, and resume execution. The Autolith project demonstrates this in practice as a self‑modifying Common Lisp agent that can inspect, edit, break, and recover its own implementation while running, using runtime features like describe and the condition system for safe self‑modification.
Shows the Autolith mascot, a stone block with a pickaxe, which represents the self-modifying Common Lisp agent discussed in the paragraph about the condition system and runtime experimentation.
Because both code and data are represented as lists, Common Lisp treats programs as data, enabling powerful macros that generate or modify code and letting developers build domain‑specific languages tailored to their problems. Henz notes her Common Lisp applications are six to seven times shorter than equivalent Python code, which reduces token costs and lets more of a program fit inside an LLM’s context window, though she acknowledges the smaller library ecosystem (Quicklisp vs. npm) and the challenge of finding engineers, suggesting that LLMs can write missing parts and that hiring learners mitigates the talent shortage.