Field Notes · Engineering & Policy

Writing for the people
who actually build this.

Most AI content on the internet is venture-funded marketing. This isn't. These are field notes from the work of building sovereign AI infrastructure for hospitals, banks, defense contractors, and government agencies — the institutions where SaaS will never be viable, and where every architectural choice has an audit consequence.

Policy 9 min read April 26, 2026

Why SaaS AI Tools Cannot Serve Regulated Buyers

Portkey, LangSmith, Helicone, OpenAI Enterprise — all built on a structural assumption that fails the moment a CISO at a Tier-1 bank or a 500-bed hospital opens the procurement form. Here is why, and what the alternative architecture actually looks like.

Read →
Cost Engineering 11 min read April 26, 2026

The True Cost of LLM Sprawl: A 12-Month Audit

A walkthrough of the cost decomposition that surfaces in every enterprise we audit. Where the money actually goes, why it grows 5–8× faster than usage, and the eight controls that flatten the curve without throttling the work.

Read →
Engineering Reference 14 min read April 26, 2026

Building HIPAA-Compliant AI Pipelines: An Engineering Reference

A line-by-line walkthrough of the architecture you actually need: PHI handling, BAA scope, audit trail granularity, encryption boundaries, and the specific 45 CFR § 164 sub-parts your build will be measured against. With a reference deployment and the code that implements it.

Read →