AI
AI systems, agentic workflows, and the model ecosystem
5 postsHarness Engineering: Why the System Around the Model Matters More Than the Model
The model is 10% of the problem. The other 90% is orchestration, guardrails, memory, and context management.
ArchitectureLLM
Context Engineering is the New Prompt Engineering
Prompt engineering was the appetizer. Context engineering (what you feed, when, and how) is the main course.
PromptingRAG
Use TypeScript Types, Not JSON, for Schema Output
Why JSON Schema is the wrong abstraction for structured LLM output, and how TypeScript types do it better.
LLMTypeScript
The Power of Mermaid Charts in Prompts
How embedding mermaid diagrams in prompts dramatically improves LLM reasoning on complex workflows.
PromptingVisualization
Edge Cases in Chunking That Will Break Your RAG Pipeline
Tables, headers, nested lists, and code blocks. The chunking strategies that look great on demos and break in production.
RAGProduction