Stop Prompting — Greg Pstrucha, Sentry
You correct your agent. It says "You're absolutely right." Five minutes later, after compaction, it makes the same mistake again.
Greg Pstrucha, who works on the infrastructure behind Seer, Sentry's debugging agent, argues you should stop re-prompting your coding agent and start codifying the rules you keep repeating. Agents have no memory, so the same elementary mistakes come back: overly defensive code, useless tests, endpoints nobody needs. Tests, strict types and especially linters turn those corrections into deterministic checks, and with agents, writing a custom lint rule is now cheap. He shows real examples from Sentry: lint rules that keep endpoints in sync with the OpenAPI schema, and a long game of whack-a-mole to keep agent skills accurate. For qualitative judgment, he uses written policies and a "Grumpy Engineer" review loop, and warns that agents will game any number you hand them, like test coverage.
The text is the source's own description of its publication. The content belongs to AI Engineer.
More from AI Engineer
Inside Arca's Long-Horizon Agents That Run the Business - Aman Singh, Arca
AI Engineer 23 Oct
Real AI Agent Attacks and What Would Have Stopped Them — Andrey Rybka, Bloomberg
AI Engineer 23 Oct
Outside of tech there has been no financial payoff to US companies for using AI, now viewed as a margin-hurting cost center, as VC-backed firms burn more tokens
Techmeme 4 h ago
Source: Replit recently moved most AI spend from Anthropic to OpenAI, which gave 30%+ off in a deal worth $100M+/year, after Anthropic discontinued its discount
Techmeme 6 h ago