1 article from the team building this since 2011.
Traditional QA asserts that output equals expected. Agentic systems produce a different sentence every run and can still be correct — or produce the same sentence twice and be wrong both times. The assertions have to move.
3 min read · Manu Sunny
