Field notes from the work
Ideas worth testing in the real world.
Practical writing about what current AI can do, where it still struggles, and the systems teams need around it.
99.9% of People Have No Idea What AI Can Do Now
AI capability has moved from answering questions to completing long, tool-using work. Most people and most organizations are still operating from an older picture.
Read the field noteSystems · 11 minDemos Are Easy. Systems Are Not.
A model response can be impressive in five minutes. Dependable work requires context, tools, permissions, evaluation, recovery, and an owner.
Read the field noteDecision framework · 10 minAgent, Automation, or Ordinary Software?
A practical framework for choosing the least complicated system that can reliably complete the work.
Read the field noteData · 11 minMost AI Projects Are Data Projects First
The model is rarely the first hard problem. Access, meaning, permissions, freshness, and feedback determine whether the system can work.
Read the field noteLeverage · 11 minWhat One Person Can Build Now
AI does not eliminate expertise. It gives a capable person enough execution leverage to cross boundaries that once required a small organization.
Read the field noteArchitecture · 12 minThe Real AI Stack: Models, Context, Tools, and Evaluation
The model creates capability. Context, tools, state, controls, and evaluation determine whether that capability becomes a dependable product.
Read the field noteDecision framework · 10 minBuild, Buy, or Integrate?
A decision framework for acquiring AI capability without buying a platform where a feature would do, or building a platform where a product already exists.
Read the field noteEvaluation · 12 minHow to Know Whether an AI Agent Actually Works
Evaluate the completed job, the decisions inside it, and the operating cost. Do not accept a confident final message as evidence.
Read the field note