# Short checked segments

Cut a long task into short segments and check each one.

**Is a:** [Guard](/reference/ontology/unattended-agent-failures/guard)

Our own conclusion, not a source's: if the failure rate is constant, success falls with task length, so shortening each unchecked stretch is the lever. The constant-rate model is one analyst's fit ([Ord, half-life of AI agent success](https://arxiv.org/abs/2505.05115)).

---

Published 2026-10-29. Web page: https://www.mackenziebowes.com/reference/ontology/unattended-agent-failures/short-checked-segments