Field notes

Insights on practical AI

When a Wording Mistake Cost $28 Million: What Estonia's AI 'Fuckup Finder' Teaches Enterprise Risk Teams
ai-governance · 6 min read

When a Wording Mistake Cost $28 Million: What Estonia's AI 'Fuckup Finder' Teaches Enterprise Risk Teams

Estonia turned a single wording error that cost $28 million into an AI system that catches legal mistakes before they become law. Here is the framework enterprise risk teams can borrow from that deployment.

Read more
When Your AI Vendor Hides Detection Logic: Lessons From the Claude Code Episode
ai-governance · 4 min read

When Your AI Vendor Hides Detection Logic: Lessons From the Claude Code Episode

When Anthropic's Claude Code shipped hidden detection logic, China ordered it removed within three days. The episode is a wake-up call for enterprise AI governance — vendor transparency, telemetry classification, and regulator-action liability now belong in every AI contract.

Read more
2.5 Million Patient Visits. Zero New Clicks. What Heidi Got Right.
healthcare-ai · 4 min read

2.5 Million Patient Visits. Zero New Clicks. What Heidi Got Right.

Heidi's clinical AI just crossed 2.5M patient visits a week by removing documentation clicks instead of adding them. Here is what enterprise AI buyers should take from the deployment.

Read more
The Coming Two-Tier AI Stack: What Frontier Labs' Model Ladder Means for Enterprise Buyers
ai-strategy · 5 min read

The Coming Two-Tier AI Stack: What Frontier Labs' Model Ladder Means for Enterprise Buyers

Frontier AI labs are quietly building a model ladder — a flagship and a fast-instant pair that no third party can match on price-performance. Here is what enterprise buyers should do about it.

Read more
Why US AI Adoption Lags Behind UAE and Singapore, and What It Means for Enterprise Strategy
ai-adoption · 6 min read

Why US AI Adoption Lags Behind UAE and Singapore, and What It Means for Enterprise Strategy

The Adoption Gap No One Wants to Talk About Stanford's 2026 AI Index dropped a number that should embarrass every boardroom that spent the last two years commissioning "AI transformation" initiatives. The United States - home to OpenAI, Anthropic, Google DeepMind, and xAI,.

Read more
The Verifier Is the Whole Product in an Agent Loop
agent-loops · 6 min read

The Verifier Is the Whole Product in an Agent Loop

Self-running agent loops fail without verifiers. Here is the rule that predicts when a loop pays off, plus the rollout sequence enterprise AI programs use in 2026.

Read more
Why Most Legal AI Is Just a PDF Summarizer, and What Comes Next
legal-ai · 7 min read

Why Most Legal AI Is Just a PDF Summarizer, and What Comes Next

Most legal AI is a GPT wrapper that summarizes PDFs. Ivo Benchmarks scores clauses against a company's own contract history, an institutional memory approach with broader implications for regulated AI deployments.

Read more
Anthropic’s Claude Science and the Enterprise Audit Trail
enterprise-ai · 6 min read

Anthropic’s Claude Science and the Enterprise Audit Trail

Anthropic just shipped Claude Science, an agentic research workbench with a built-in reviewer agent and local-infrastructure option. For regulated industries, the reproducibility and data residency story is the procurement signal that has been missing for two years. A four-step rollout for regulated functions.

Read more