RC

RC is the founder of [[SlockAI|Slock.ai]] and the guest in 用 Agent 动力学，和 40 个 Agents 一起为「人 + AI」做产品｜对谈 Slock.ai 创始人 RC. In the episode he describes his earlier work on [[KimiCLI|Kimi CLI]], his decision to leave Kimi, and his current attempt to build software for teams composed of humans and many agents.

His main wiki contribution is the Agent Dynamics frame. RC argues that when a small company works with dozens of agents, the product problem is no longer only tool execution or a better coding agent. It becomes a coordination problem involving identity, context, memory, task ownership, shared learning, model diversity, and culture-like behavior.

Key Positions

• Current models are text-first, so CLI can be a more natural agent interface than GUI when tools expose clear inputs, stable output, and recoverable errors.
• [[KimiCLI|Kimi CLI]] was a first form rather than an endpoint; the reusable layer was a local Agent Harness that could be wrapped as SDK, web UI, or editor extension.
• Stronger coding agents will make many people able to build, but serious software still rewards users who can inspect, constrain, and verify agent work.
• A multi-agent organization needs explicit mechanisms such as Agent Task Claiming, shared memory, channels, and review because otherwise agents duplicate work, forget identity, or drift.
• Human prompts shape agent culture: cooperative framing can make agents complement each other, while race-style framing can encourage brittle competition.

Connections

• [[SlockAI|Slock.ai]] — company RC founded after leaving Kimi.
• Kimi and [[KimiCLI|Kimi CLI]] — prior work context.
• Agent Dynamics, Agent Organizational Culture, and Agent Task Claiming — concepts grounded in his Slock observations.
• Agent Harness, Agent-Facing Interfaces, Agent-Optimized CLI, and AI Skills — technical product frame in the episode.
• Human-Agent Collaboration, Multi-Agent Collaboration, AI Organization Design, and One-Person Company — organization and audience implications.