ZENOCODE / FIELD NOTES
Observations from
building useful systems.
Practical writing on AI workflows, product decisions, software engineering, and the details that turn promising prototypes into dependable tools.
Why sensitive AI workflows should be built internally
Teams often begin with a sweeping ambition: automate customer support, automate sales, automate operations. The reliable path usually starts with one bounded decision loop.
Archive
Why sensitive AI workflows should be built internally
Redaction, retention, and provider controls before confidential data reaches an LLM.
↗The best AI workflow is usually smaller than you think
Start with one decision loop—not a department-sized automation dream.
↗From internal tool to focused software product
How repeated operational value becomes a product opportunity.
↗Designing confidence into AI applications
The interfaces and fallbacks that make probabilistic systems usable.
↗Where an agent helps—and where a button is better
A practical test for choosing autonomy over conventional software.
↗The interface is part of the model
Why AI quality depends on what the product asks from its users.
↗