Writing Lisp is AI resistant and I'm sad
blog.djhaskin.com
April 5, 2026
6 min read
🔥🔥🔥🔥🔥
50/100
Summary
AI systems show a preference for programming languages other than Lisp, making it challenging to use Lisp for certain tasks. A developer is creating a tool to convert between different RSS reader formats, initially using Lisp, but faces difficulties in training the AI to understand the language.
Key Takeaways
- AI struggles significantly with Lisp programming, leading to inefficient development and higher costs compared to languages like Python.
- The author created a tool called tmux-repl-mcp to simplify REPL interactions for AI, but it still resulted in high expenses and limited progress.
- The high-latency nature of AI API interactions is incompatible with the low-latency benefits of REPL development, making Lisp particularly resistant to effective AI assistance.
- The author is considering rewriting their project in Go due to better AI support and training data availability for that language.
Community Sentiment
Positives
- AI models like Codex and Claude Code excel at writing idiomatic Clojure code, demonstrating that Lisp-like languages can be effectively supported by AI.
- LLMs can handle simpler programming languages well, suggesting that their performance is heavily influenced by the complexity of the language and its training data.
- Users report success in using AI for Common Lisp projects, indicating that with the right prompts, AI can assist effectively in more complex coding environments.
Concerns
- The article's claim that writing Lisp is AI resistant overlooks the capabilities of AI in handling languages like Clojure and Scheme.
- Complexity in languages like GCL poses challenges for AI, suggesting that not all programming languages are equally accessible to AI models.
