Connecting production traces to experiments
Braintrust’s workflow collects application traces, analyzes logs, adds human feedback to datasets and compares changes in experiments. It also supports scoring production traffic. This connects investigation of a real failure with a repeatable test, while the choice of dataset and scorer still determines what the evaluation measures. [1]
An evaluation product introduced in beta
The September 2023 introduction presented versioned datasets, a comparison interface and a prompt playground. Initial access involved beta customers; it should not be read as a general commercial launch. [2][3]
Sources
- Get started with Braintrust
Braintrust · retrieved Oct 23, 2026
- It’s time to build reliable AI
Braintrust · Sep 28, 2023
- The hot new thing: AI platforms that stop AI’s mistakes before production
TechCrunch · Sep 28, 2023