Notes from production.
Long-form essays on agentic systems, distributed-systems thinking, product craft for engineers, and self-hosting LLMs on GPUs. Opinions are mine, evidence is cited.
- 01
Practical vs Theoretical Agentic Systems
Most production wins called "agents" are workflows with one tool-calling node. A field report on what actually ships, where the loops break, and why MCP — not the planner — was the real unlock.
Agentic AILLMsMCPEVALS22 APR 2026
11 MIN READ - 02
How to Think About Systems
Notes from operating distributed systems at scale: where leverage lives, why root causes are a coping mechanism, and the architecture diagrams nobody draws but everyone runs.
SystemsArchitectureReliabilityEngineering26 APR 2026
12 MIN READ - 03
How to Think About Product
The engineer-as-tech-lead view of product thinking — why most roadmaps are theatre, how the engineer gains leverage on what to build, and the discipline of writing your "no" list in public.
ProductEngineering LeadershipStrategy30 APR 2026
10 MIN READ - 04
Deploy Your Own Models to GPUs
A 2026 operator playbook: when self-hosting wins, which inference engine to pick, the GPU buying matrix, quantization that actually works, and the four metrics you autoscale on.
LLMsInferenceGPUvLLM02 MAY 2026
16 MIN READ