Karpathy on an overlooked Pareto point in LLMs
Karpathy sees such models as a point on the LLM Pareto curve with large latent demand: no thinking, single token, low latency, acceptable intelligence, under-invested due to races for higher intelligence.