Field Notes
33 notes on this leg
Field Notes are reflections on IT leadership, AI operations, governance, security, and systems thinking — especially where technology meets human judgment.
What Is an AI Hallucination? A Plain-English Guide
An AI hallucination is a confident-looking output that invents or misstates facts. Here is how to spot and verify the…
I Was at OpenAI DevDay 2026. Dots Changed the Question From “Can AI Help?” to “What Should It Own?”
A firsthand OpenAI DevDay 2026 field report on Dots, ChatGPT Space, GPT-6.1 Sol, and the governance questions persistent AI agents…
What Is AI Red Teaming? Explained in Plain English
AI red teaming is structured adversarial testing of a model and its surrounding system. Here is what it tests, what…
OpenAI Paused Model Training After Agent Misbehavior. The Governance Lesson Is Detection.
The useful lesson from OpenAI’s training pause is not the “rogue AI” headline. It is the need for agent inventory,…
What Is AI Behavior Drift? A Plain-English Guide
AI behavior drift is an observable move away from an approved baseline. Here is how it differs from a bug,…
Anthropic’s Model Hardware Standard: How AI Agents Could Operate Physical Equipment
Anthropic's Model Hardware Standard could connect AI agents to lab and manufacturing equipment. Here are the capabilities, limits, and governance…
1,200 AI Agents, 70,000 Messages: What the UN’s New Report Actually Found
What a UN panel says a 1,200-agent incident reveals about permissions, coordination, monitoring, and recovery for agentic AI.
How to Run GPT-6 Astra Without Burning Through Your Usage Limit
GPT-6 Astra burns through usage allowance fast if you don't understand the mechanics. Here's the reset-timing trap most advice misses,…
The AI Governance Gap Nobody’s Fixing: Why 87% of Companies Use Agents but Only 47% Can Govern Them
87% of companies encourage AI agent use, but only 47% can govern it. Two vendor surveys converge on the same…
Anthropic Put Numbers on AI Building Itself: What the R&D Automation Index Actually Shows
Anthropic self-published new metrics on how much of its own AI R&D is now done by Claude. Here's what the…
