Can Multi-Modal LLMs "See" Images? A Deep Dive with ASCII Art | Dark Hacker News