Every alive subcomponent the decomposition found, paginated.

Each card is one rank-1 component pulled out of Qwen2.5-Coder-1.5B at layer 13. Auto-labeled by DeepSeek-V3 on its top firing tokens, sourced from eren23/eren-code-style. Toggle BASE / +LoRA to flip between the bare model and the personal-style fine-tune. Same matrix, same corpus, different alive sets — the diff is the fingerprint.