In the AI era, defining problems is often more important than solving them; here we explore AI's contradictions, boundaries, and possibilities.
The article examines the emerging A2A (Agent‑to‑Agent) protocol, tracing its evolution from function calling to MCP and finally A2A, and evaluates its core concepts, security model, task lifecycle, transport options, design guidelines, and operational best practices for building interoperable AI agent systems.
The article examines Model Context Protocol (MCP), introduced by Anthropic in November 2024, tracing its rapid adoption, architectural design—including Host/Client/Server roles, transport layers, security and observability practices—and outlines production guidelines, future roadmap, and current limitations.
The article analyzes how large language models process only tokenized text, compares the traditional LLM‑plus‑toolchain pipeline with emerging multimodal models, evaluates their cost, speed, controllability, and hallucination risks, and proposes a hybrid architecture that matches each approach to specific document scenarios.
The article argues that the common assumption that text chunks are the proper knowledge unit in RAG pipelines is flawed, leading to versioning, metadata, and redundancy problems, and demonstrates that replacing chunks with structured IdeaBlocks dramatically reduces corpus size, token usage, and improves vector relevance.
The article analyzes why traditional API gateways cannot meet the needs of stateful Agentic workflows and proposes a dedicated Agent gateway that handles access control, cross‑service execution tracing, and pre‑LLM security enforcement while addressing connection overhead, session fan‑out, and observability challenges.
The article analyses Markdown's minimalist design, its ambiguous syntax, security flaws such as ReDoS and XSS vulnerabilities, and the growing gap between its original simple transliteration goal and the complex compiler‑like features developers now demand.
While a single LLM call is a stateless function, real‑world tasks require dynamic information gathering, hypothesis testing, and iterative refinement, so agents must operate in a continuous loop; the article analyzes core patterns such as ReAct, Plan‑Execute, Reflection, Multi‑Agent and HITL, highlighting state management, cost, debugging, and observability challenges.
This article examines practical ways to improve Retrieval‑Augmented Generation (RAG) retrieval quality—covering vector database choices, data chunking, embedding models, query expansion, and re‑ranking—while weighing performance gains against operational costs through multiple real‑world case studies.
The article examines why pure vector retrieval in RAG lacks lexical precision and traceable relevance scores, explains BM25's strengths, and presents hybrid retrieval architectures—including RRF and linear combination fusion—as well as the trade‑offs of externalizing the fusion process.
The article reviews the three‑step RAG pipeline, explains why retrieval quality hinges on fast, accurate semantic matching, contrasts Bi-encoder’s offline vector indexing and speed with Cross-encoder’s token‑level interaction and higher precision, and discusses hybrid solutions such as ColBERT and LLM rerankers with practical engineering guidelines.
