Braintrust

You can now vibe code custom annotation UIs directly in Braintrust. Investigate specific issues, share views across your team, and debug your product in a workflow that perfectly matches your use case.

That's a wrap on re:Invent! The message from teams was consistent: switching models safely, ending manual dataset work, and getting org-wide visibility are now table stakes. Clear takeaway → agents are where the real value shows up, and AI observability is how teams ship production systems that work.

If you're at #AWSreInvent and working with Bedrock agents, learn how to deploy a Strands Agent on AgentCore with full observability. Deploy now →

The team at Retool built a data-driven workflow where production logs directly inform their prioritization decisions. Using Loop, they query production data semantically to surface insights in seconds, turning observability into a roadmap. Learn more about their workflow →

Loop helps your team move faster across the entire AI development workflow. Use Loop to create scorers, analyze logs, clean datasets, and iterate on prompts using production data. Get started with Loop ↓

To build Loop, we went deep on real conversations, patterns, and prototypes across product and design engineering. Here's how we built it →

Production AI breaks in unexpected ways. Loop helps you figure out what happened and what to fix, fast. Describe what you're looking for and Loop will help you find failure patterns, surface root causes, generate datasets, and write precise queries when you need filters. PMs use it to identify issues by impact. Engineers use it to debug and build scorers based on real traces. Teams share one view of how AI behaves in production. Learn more about Loop → https://lnkd.in/gY-TjAiA

Met with AI leaders at F1 Vegas. Clear themes from every discussion → focus, tight feedback loops, and speed.

What's new in Braintrust: - Dataset schemas and form-based editing - Publicly shareable traces - Custom monitor chart generation Read more →

Build AI that works.