Tag
agent
2 posts
- A Claude-centric software architectureFor thirty years, software has meant humans translating changing requirements into fixed code. Once a runtime like Claude Code is strong enough, the more efficient approach is to separate the two: settle deterministic logic into a data layer, a function layer and Skill methodology, and hand the uncertain requirements to the runtime to understand and render in real time. With quant-researcher as a worked example.
- Three things every AI agent engineer should know: agent architecture, prompt engineering, and MCPA platform-agnostic guide: the agentic loop and multi-agent orchestration, context passing and error propagation, few-shot and structured output, controlling false positives, and the structure, message format and a minimal server implementation of the Model Context Protocol.