research
Research
Architecture analyses, engineering notes, benchmarks and postmortems — what we learned in practice, not text written for search rankings.
Agent governance: why AI assistants don't scale without control
What happens when an autonomous agent is allowed to grow without authority boundaries or auditing, and how governance is built into the architecture rather than bolted on after an incident.
AI Economics: why the token bill grows unnoticed
Where uncontrolled AI-system cost comes from in production, how to measure it per workflow step, and which decisions actually cut spend without losing quality.
Context engineering: why RAG breaks in production
Why knowledge-retrieval systems look perfect in a demo and degrade within a month, and which engineering decisions a working RAG is actually made of.
LLM orchestration: why wrappers don't survive the second version
How a working AI system differs from an LLM wrapper, where wrappers break under real requirements, and what orchestrating a workflow around the model is actually made of.