Blog
Research & perspectives
PerspectiveJuly 2026
The correction loop is the bottleneckThe path from an expert's correction to a shipped change in agent behavior (a spreadsheet of traces, a Jira ticket, a barely tested prompt edit) is the real constraint on enterprise agent deployment. Why that loop exists, why it's the bottleneck, and what replaces it.Read →
ResearchJune 2026
Specifying the agent: closing the gap to frontier with a context engine and meta-distillationSearch and learning algorithms that attack specification failure directly (a context engine and meta-distillation), evaluated across the τ³-bench banking suite, FinanceBench, and the Harvey Legal Agent Benchmark on two open / open-weight backbones.Read →