June 9, 2026
Why I'm Writing About LLM Systems and Agents
A first note on building a public learning practice around LLM systems, agents, evaluation, and the engineering habits behind them.
- LLMs
- Agents
- AI Engineering
- Writing
I am starting this blog as a public notebook for learning how modern AI systems are built, tested, and improved.
The center of gravity will be LLM systems and agents: tool use, memory, retrieval, evaluation, orchestration, product behavior, and the strange gap between a promising demo and a reliable system.
I want the writing here to be useful in three ways:
- It should make my own thinking harder to fake.
- It should leave behind artifacts that other builders can inspect, reuse, or challenge.
- It should turn fast-moving AI research into practical engineering judgment.
The format will vary. Some posts will be implementation notes. Some will be paper-to-code walkthroughs. Some will be failure analyses. Some will be opinionated essays, but only when the opinion has earned its oxygen through examples, measurements, or lived engineering friction.
The long-term bet is simple: a strong technical voice is built by repeatedly producing clear artifacts in public.
This is the first one.