Show HN: LLM Rescuer – Fixing the billion dollar mistake in Ruby (github.com)

LLM Rescuer is an experimental Ruby gem that monkey-patches NilClass to intercept NoMethodError calls on nil objects and asks an LLM (the project jokes about “GPT‑5”) to guess what the developer probably meant, returning a generated value so execution continues instead of crashing. Setup is simple—add gem 'llm_rescuer', provide an OPENAI_API_KEY, configure a project prefix and call LlmRescuer.setup—but it relies on dependencies like ruby_llm, ruby_llm-schema and binding_of_caller to capture context. The repo is explicitly a proof‑of‑concept and warns about unpredictable behavior, security risks, and API costs (est. $0.002 per rescue and wide-ranging monthly bills). For the AI/ML community this is an intriguing, cheeky demonstration of LLMs used for runtime repair and contextual code completion at execution time: it highlights how models can be applied to hot‑patch behavior, do speculative program synthesis, and autocomplete missing values based on surrounding code. But its technical implications are serious—non‑determinism, hallucinations, leaking code/context to an external API, testing and debugging difficulty, and the dangers of monkey‑patching core classes in production. As a research toy it surfaces interesting questions about LLM reliability for self‑healing systems; as a tool, it’s a risky anti‑pattern not suitable for production without heavy constraints.