Inference costPublished - April 9, 2026
Batch Inference
Why batch inference can cut costs sharply on AI workloads that do not need real-time latency.
Value often starts with better cost architecture, not with a bigger model.
Lab
Public posts and projects
Articles, repos, and working notes, openly available.
Agentic workflowsPublished - January 27, 2026
Claude-Book
The current public Claude-Book release shows how I orchestrate multiple agents, state, and multi-pass workflows around a writing system.
How to design agentic workflows that go beyond a wrapper or a linear chatbot.
RAG simplificationPublished - March 29, 2026
retrieval-arena
A repo to compare retrieval strategies and document when embeddings actually add value.
Reducing RAG stack complexity often improves delivery speed, maintenance, and total cost.