Are unit tests even necessary anymore (under agentic development)?
retained text
While working in AI-driven repos, I'm wondering if its even worth having the AI write unit tests. Pre-AI, they were non-negotiable guardrails. But now with AI doing all the programming, like - what's the point? In agentic dev, entire modules get rewritten in minutes. Writing granular unit tests for code that changes every three iterations feels like building scaffolding for a sandcastle. Most real-world bugs are architectural or integration issues that 10-line mocks never catch anyway. And anyway I almost never see unit tests catch something that could have been a big problem. Theres usually a growing tests folder consuming tokens but they just about always pass.
document fetched 2026-11-11 00:40 · canonical https://reddit.com/r/ClaudeCode/comments/1xsvehx/are_unit_tests_even_necessary_anymore_under