Show HN: Neurolume – Fine-tune LLMs on a Mac and watch what changes per layer
Category: ai-ml
Tags: fine-tuning, lora, interpretability, macos, mlx, llm, local-first
Score: 7.3/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 6)
Neurolume is a local Mac app that combines LoRA fine-tuning of LLMs with real-time layer-level interpretability: a 3D 'brain' view, per-layer activity, logit lens, per-token confidence, and before/after comparison. It stands out by making training and introspection a single workflow, with careful honesty about which signals are real vs. stylized.
Target audience: AI/ML engineers, researchers, and advanced hobbyists working with local LLMs on Apple Silicon
Repository: https://github.com/Yiteng-CHEN/neurolume · Python · MIT
View on Hacker News