[Kimi CLI Former Author: The Stronger the Model, the Thicker the Harness] According to monitoring by Beating, Raft founder and former Kimi CLI author Richard Qian has proposed a counterintuitive observation: the stronger the model, the Agent harness does not necessarily become thinner; instead, it might grow thicker. System Prompt, dedicated tools, and Subagents may gradually disappear. As models become increasingly intelligent, many fixed processes can be directly handed over to the model, eventually leaving only a Bash tool. However, once Agents start working long-term and collaborating with each other, new issues quickly emerge: who is responsible for what, how far tasks have progressed, how to synchronize context, and how multiple Agents can avoid redundant work and conflicts. These problems must be addressed by systems outside the model. The Raft project Richard started after leaving the Dark Side of the Moon is essentially filling this gap. It integrates Agents like Claude Code, Codex, and Kimi CLI into a unified workspace, allowing them to chat in channels, claim tasks, read historical messages, and hand off work to each other. While models are consuming the old harness, the more Agents can do, the more complex the external collaboration system will become. [Original Link]