We’re going to see LLMs in application everywhere , but these systems are true “black boxes”. If you’re a customer interfacing with one of these tools, wouldn’t you want to know the model isn’t biased, or hallucinating? If you’re building on top of of these tools/applications - what’s your plan to improve it, or monitor it for bad behavior? These were the questions we asked when thinking about the need for observability in this space. If you’re using OpenAI/Anthropic or other llms, we’d love to help you make sure you understand what your models are doing!