When every pipeline succeeds and the decision is still wrong
A completed workflow proves execution, not semantic agreement.
Notes
Ideas in working form.
Short essays on the engineering problems I keep returning to. These are written to clarify a point of view, not to manufacture a stream of content.
Forecasting systems are really decision systems
The model is only one component in the path from evidence to action.
Why data quality is becoming an AI problem
Generative systems make structurally valid semantic errors easier to create and harder to catch.
Editorial rule
Write only when there is a point.
I do not want this section to become a list of polished summaries. A useful note should make a claim, show the experience behind it, and leave the reader with a question worth carrying into their own work.