According to @DeepLearningAI, Andrew Ng argues that LLMs are general but limited, with progress currently driven by piecemeal, data-centric, domain-specific work and no quick leap to AGI, as outlined in The Batch for Dec 18, 2025 (Source: DeepLearning.AI, The Batch, hubs.la/Q03YD9Tx0). The update also flags four trade-relevant headlines: Runway’s GWM-1 enabling real-time, controllable world-model video; Disney teaming up with OpenAI; OpenAI rolling out a GPT-5.2 suite; and researchers unveiling SEMI, a method that teaches LLMs new data types with roughly 32 examples (Source: DeepLearning.AI, The Batch, hubs.la/Q03YD9Tx0). The communication includes no cryptocurrency items, indicating any crypto exposure here is indirect via broader AI sentiment rather than token-specific news (Source: DeepLearning.AI, The Batch, hubs.la/Q03YD9Tx0).