The integration of AI and cryptocurrency is accelerating, having gone through two stages, and the third stage is now taking shape.

Looking back at the first two steps: tokenizing computing power was the earliest, with projects like Render packaging GPU computing resources into tradable tokens to facilitate idle compute power flow. Next came the exploration of model tokenization, with Bittensor transforming AI models into incentive-driven networks, allowing training and inference processes to be quantified and rewarded.

Where is the new hotspot now? It lies in the behavior and value flow of tokenized AI. Simply put, it’s about enabling every decision, every interaction, and every value creation by AI to be tracked, incentivized, and exchanged. This is the core opportunity of AI agent infrastructure—building a complete economic system where AI is no longer a passive tool but an active participant in the economic network. The imaginative space at this level is much larger than the previous two steps.