I saw this reinforcement learning in OpenAI's latest published paper ➕ I was stunned when calculating the model during testing! Isn't this the instantiation of the AI Agent → AGI technology roadmap pie drawn by AI16Z's joint creation shaw when DeepSeek caused the AI Agent crash a while ago? Now OpenAI has internalized the Agent into the LLM inference training layer and used large-scale computing power to optimize the parameter weights of the Agent network. Previously popular AI Agent projects could only sigh helplessly at this point. The verification of the OpenAI AGI model implies the falsification of the AI16Z narrative. If the AI Agent in the encryption field can only capture the value of the application layer niche market (DeFAI), the ongoing de foam process will be difficult to end. In an optimistic situation, we still have to wait until LLM Scale When Law encountered another bottleneck, A16Z acted as a white horse rider to invent a new AI x Crypto narrative.