Blog

AI observability is active observability

1 June 2026Ankur Goyal2 min

Today, Topics is GA. On the surface, it looks simple. It assigns a few short summaries to every trace, then clusters them automatically. Isn't that a solved problem?

In an industry flooded with futuristic workflow promises, we've been fascinated by a more fundamental problem: building automation that can manage the volume of traces our customers generate, in a way that doesn't blow out token costs or break down at serious scale.

Even something as simple as topic clustering hasn't worked at the scale of millions of traces per day. We sat down with customers like Replit and decomposed this problem into every piece: how do you reconstruct millions of conversational threads every hour, run an inexpensive yet powerful LLM on them, store vectors so you can cluster them on-demand, and expose all of this in a simple UI?

In doing so, we built the foundation for the next chapter of Braintrust, which is about cost-effectively applying intelligence at scale to your traces. Our thesis is that it's too expensive to expose your coding agent to 100% of your production data. Our job is to ship tools that help you distill that data into the relevant handful of things it should look at.

Braintrust now works on this continuously, in the background, and gives you the primitives you need to derive deeper insight, both through SQL and our UI.

We think of this next chapter as moving from "AI observability" to "active observability". We work behind the scenes, constantly, to find answers to questions without you having to ask them ad-hoc.

The way we build software is changing quickly, and a few clear layers of the stack are emerging. But Braintrust shouldn't lock customers in to a specific vision of what that stack looks like. Our goal is to enable you to utilize frontier intelligence, cost-effectively, and distill your production traffic into useful information you and your agents can use to build quality products.


Want to see what Topics finds in your production data? Get started with Braintrust or book a demo.

Share

Trace everything