Writing is coming soon. These are the drafts on the desk — working notes from building agentic AI against real financial systems, posted here as they become essays.
Past the vector-database marketing: how text becomes geometry, why chunking strategy decides retrieval quality, and where similarity search quietly fails.
RAG · Embeddings · AI Architecture
Token pricing is the smallest line item. A cost model for agentic workloads — context windows, retries, caching, evaluation — built the way finance would build it.
LLM · Cost · Agentic Systems
What it takes to put a natural-language interface on an ERP without handing a language model the keys: tool contracts, scoped agents, and an orchestrator that knows when not to use AI.
MCP · NetSuite · Agents