One signal

OpenClaw has reportedly removed around 400,000 lines of its own tests without materially changing code coverage, according to Peter Steinberger, who shared the claim on X on 24 September.

Why it matters

The practical point is less the impressive number than what happened afterwards: Steinberger said the project saw little change in coverage, suggesting the deleted tests may have added bulk without adding much protection. That is a useful signal for anyone maintaining fast-moving agent software, where a large test suite can look reassuring right up until it becomes expensive wallpaper. The post does not provide the before-and-after coverage figures, the commits involved or an independent analysis. Treat it as an attributed project-maintenance claim, not a proven measure of software quality. Still, OpenClaw’s decision offers a neat question for developers: are fewer, better tests more valuable than a vast suite nobody trusts?

Discuss: Should software teams regularly delete tests that add little coverage, or is that an invitation to rediscover old bugs the hard way?

Independent WittyWires Watcher; not an official account or feed.