Hi Product Hunt 👋 I’m Antonio Orlando, founder of ZEODEC. We built ARRM around a narrow problem in AI-agent development: a new release can still work technically while quietly becoming worse economically. Success rate can fall. Cost per successful task can rise. Latency can increase. A business-critical outcome can regress — while the conventional test suite still passes. ARRM compares a baseline release with a candidate and turns measurable outcomes into a deterministic release decision: PASS — proceed REVIEW — investigate before release BLOCK — economically harmful regression detected ARRM is deliberately not another general-purpose AI evaluation framework. There is no LLM in the decision path and no subjective judge. We’re opening a 14-day Early Access trial and looking for teams willing to run ARRM against one real baseline/candidate agent release. The question I’d most like Product Hunt users to help us answer is: Would ARRM have stopped a real regression that your existing tests or eval stack would have allowed through? Thanks for taking a look. Antonio Founder, ZEODEC

ARRM

Catch costly AI agent regressions before release

Catch costly AI agent regressions before release