Talk about your next project
Braintrust case study

The lesson

Open with customer proof and a clear product problem, then show the testing system that made the result possible.

The move

The page starts with the Bubble logo, the title “How Bubble turns user prompts into production apps,” named Bubble engineering leaders, and a customer video. Directly below, three proof points summarize scale and outcome: “6M apps built on Bubble,” “1.5x more accurate Agent after evals,” and “2 automations triaging traces and running nightly evals.” The story then explains Bubble’s AI app builder, moves into “The limits of manual testing,” compares “Before Braintrust” and “After Braintrust,” and follows with sections on defining correctness, inspecting traces, shipping Agent 2, and “Key takeaways.”

Why it works

  • Our readThe first screen gives the buyer the customer, the product use case, the named speakers, and the video before asking them to read.
  • Our readThe before-and-after table makes the operational change easy to scan, from manual testing and repeated prompts to datasets, scorers, tracked tradeoffs, nightly evals, and auto-triage.
  • Our readThe story stays close to the buyer’s real evaluation work by naming user intent, equivalent outputs, trace inspection, latency, cost, and model comparison.

What not to copy

Don't copy Bubble's Agent, no-code history, app-building prompts, or Claude-driven automation details. Keep the structure of video-led customer proof, problem context, before-and-after workflow, technical method, launch outcome, and takeaways.

More from Braintrust

More case studies

Steal it for your brand.

Tell us what you want to make. We’ll talk through how to put these ideas to work for your brand.