Interview with Kevin Ding of Huasi Dynamics: When choosing an agent for the future, it’s important to see what it has learned in the past thirty days.
Kevin Ding, founder of Pyromind Dynamics, believes that the key to Agent competition in the future lies in their ability to continuously learn from work results, rather than just relying on basic model capabilities. This view stems from his observation in GUI Agent experiments of the bottleneck where existing Agents “can work but cannot grow”. Currently, both Salesforce and OpenAI emphasize the importance of continuous improvement after launch, and Pyromind Dynamics has taken the lead in productizing Agent continuous learning. Kevin Ding points out that as basic models become more mature, production feedback data accumulates, and reinforcement learning infrastructure improves, market demand and technical conditions are aligned for the first time, prompting companies to shift from “one-time delivery” to a new model of “starting learning immediately after launch”.