Post

Maina Wycliffe

@mainawycliffe.dev

share.googleHow We Benchmark Deep AgentsWe revamped how we benchmark Deep Agents. Here's the eval setup we run in Harbor across coding, conversation, and retrieval, and how we use it to ship changes.

27/7/26, 3:06 pm·0 reposts·0 likes