Stop Prompting — Greg Pstrucha, Sentry
Replace repeated agent prompting with codified rules via tests, typing, and linters
“you should stop prompting, you should start codifying the actual rules that the agent should follow”
Sentry AI engineer Greg Pstrucha argues that prompting coding agents to follow standards is futile because context compression causes agents to revert to bad habits after fixes. His solution: codify expectations as deterministic checks — tests, strong typing, and linters — so agents operate within hard constraints rather than soft conversational guidance. This is a practical shift in how engineering teams should think about AI agent governance at the code level.