01
AGENT ENGINEERING
Agent Engineering
Harness design, tool orchestration, multi-agent systems — where the 98% of engineering beyond the model actually lives.
Agent Engineering column
AI Agent · Field Notes
Building with Intelligent Agents
Field notes from Xinwei Xiong (cubxxw) on taking large models out of the chat box and into production systems. The work spans three directions — agent engineering, context engineering, and Generative Engine Optimization (GEO) — all drawn from real builds, real failures, and real retrospectives.
02
CONTEXT ENGINEERING
Context Engineering
Context, memory, RAG and retrieval — what bounds an agent's intelligence isn't the model, it's what the model can see.
Start with this one03
GEO
Generative Engine Optimization
When search shifts from giving links to giving answers, visibility is decided by the probability of being cited by AI.
GEO series
Latest Signal
Nango in Practice: Let Users Add Their Own MCP Servers in Chat
Connect user-provided MCP servers to a chat product with Nango. Run a project-status demo, follow the authorization flow, and fix five production pitfalls.
Series
All columns
SERIES
01 / 07
Personal Agent Studies
Open column
SERIES
02 / 07
Agent System Design AnatomyControl, state, identity, and responsibility in real systemsSource-grounded studies of Pi, n8n, and OpenClaw: trace who controls the loop, owns persistent …
Open column
SERIES
03 / 07
Agent EngineeringFive essays from context foundations to production systemsFive interlocking long-form essays: the foundations of context engineering, a panoramic map of agent …
Open column
SERIES
04 / 07
The Super Individual's Gear StackYour gear is arming your competitors tooA five-essay column. Almost every "super individual gear list for the AI era" answers the same …
Open column
SERIES
05 / 07
From Information to CreationIn the AI era, I split my notes into four different thingsAn ongoing column: most people treat information, records, knowledge, and creation as the same …
Open column
SERIES
06 / 07
AI 2026: First-Half Review, Second-Half ForecastFive essays — from what actually happened, to what comes nextA five-part column of observation and forecast. I take the things that actually happened in AI Agent …
Open column
SERIES
07 / 07
GEO · Generative Engine OptimizationWhen search stops giving links and starts giving answersAn ongoing column: when AI search engines answer directly instead of linking out, the visibility …
Open column
All Articles
68FAQ
What is Agent Engineering?
Agent Engineering is the discipline of building production-grade AI agent systems. Model capability is only a small fraction of the work — most of the engineering lives in the harness (execution framework), tool orchestration, context management, memory systems, and evaluation. The "Agent Engineering Panorama" article in this section maps these layers systematically.
How is Context Engineering different from prompt engineering?
Prompt engineering optimizes the wording of a single instruction; context engineering manages everything the model can see across an entire session — retrieval results, memory, tool outputs, history compression. As long-running agents become the norm, context engineering is replacing prompt engineering as the new foundation of AI applications.
What is GEO (Generative Engine Optimization)?
GEO is the practice of making content easier for AI search engines like ChatGPT Search and Perplexity to retrieve and cite. Unlike traditional SEO, which optimizes for click-through ranking, GEO optimizes for the probability of being cited by AI. This section carries a six-part series covering GEO from principles and structured-content tactics to measurement tooling.
What experience is this section based on?
Everything comes from the author's (Xinwei Xiong / cubxxw) first-hand practice building AI products — open-source projects, production agent deployments, and measured data from this blog's own GEO rebuild — not second-hand summaries. Articles are organized into series; start from part one of any series.