Evolve the Harness, Then the Model

As AI agents take on longer and more complex workflows, an important question is emerging: how can agent systems learn from experience and improve over time? Building that kind of learning loop involves more than the model itself. It brings together the agent environment, tools, memory, feedback, evaluation, and post-training—and requires careful choices about how these components work together. Join Hermes × AReno × Ant Ling for a practical conversation on evolving agent systems with Ling-3.0-tiny. Bringing together perspectives across the agent, model, and training ecosystems, we’ll discuss the technologies, design considerations, and engineering trade-offs behind building more adaptive agent systems. Whether you’re working on agents, models, training, or AI infrastructure, this session offers a closer look at how agent learning and self-improvement are taking shape in practice.