Personal writing on technology, architecture, AI, digital trust, and the occasional Friday coffee.
Two failures from real builds show two different reasons AI gets enterprise software wrong: one answer was never written down, the other was — but lost to the volume of outdated material saying something else.
A coding agent can write good SQL. It cannot remember what it already ran. sql-runner gives Oracle Database projects an inspectable engineering history — built out of text files, not embeddings.
Agents removed the assumption that the application would always be in the loop. What remains trustworthy when the caller is not?
Two leading language models pulled a white paper in opposite directions. The useful workflow uses both, then puts human judgement back in charge.