LLM X/Twitter 1h ago 1 min read Karpathy’s Opus 5 test turns a 1M-token prompt into 5,500 lines LLM evaluation is moving from short visual prompts to long-running code generation. Karpathy says Opus 5 spent about two hours and a 1M-token budget, roughly $10, to produce 5,500 lines of Three.js. #karpathy#opus-5#multimodal