About This Blog
An introduction to this blog's scope and editorial approach—technical, citation-oriented writing on AI orchestration, agent frameworks, and LLM infrastructure.
This blog publishes technical articles on the infrastructure underlying LLM-based systems: orchestration frameworks, agent-coordination patterns, memory systems, observability, inference serving, and the reliability properties that distinguish production systems from prototypes.
Editorial approach
The articles are written in a technical, citation-oriented register. We aim to:
- Attribute claims to primary sources. Empirical and quantitative statements are linked to papers, official documentation, or primary announcements. Figures that cannot be so attributed are omitted or explicitly qualified.
- Distinguish evidence from opinion. Vendor-reported benchmarks are labelled as such, and analytical claims are separated from established results.
- Favour comparison over advocacy. Where multiple systems address the same problem, we characterise their design trade-offs rather than rank them categorically.
Scope
Recurring topics include:
- Orchestration frameworks — LangGraph, CrewAI, AutoGen, Haystack, and the Microsoft Agent Framework.
- Memory systems — Mem0, Letta, Zep, and the benchmarks used to evaluate long-term memory.
- Observability — tracing, evaluation, and monitoring for non-deterministic systems.
- Inference serving — vLLM, SGLang, TGI, and the algorithms behind them.
- Durable execution — reliability patterns for long-running agentic workflows.
Each article includes a numbered reference list so that readers can consult the underlying sources directly.