When companies deploy artificial intelligence without observability in place, they often have little visibility into what their models or agents are actually doing under the hood.

Honeycomb.io’s platform is designed to address that gap in understanding. Using OpenTelemetry as its open-source framework, Honeycomb provides users with a way to visualize and analyze their telemetry data.

Honeycomb.io’s Graham Siener talks about the need for AI telemetry.

“People are consolidating,” said Graham Siener (pictured), vice president of product at Honeycomb (Hound Technology, Inc.). “They’re using a lot of different tools, and they’re realizing it’s actually more valuable if they have all their data in one spot. The macro trend we’re seeing in observability is people are looking at all these different applications, they’re looking at all this telemetry and they’re saying it needs to connect together.”

Siener spoke with theCUBE’s Rob Strechay at the KubeCon + CloudNativeCon NA event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed Honeycomb’s observability platform and rebuilding legacy ecosystems around AI. (* Disclosure below.)

Observability on AI activities

When faced with integrating AI into legacy applications, companies are rebuilding them from the ground up instead, according to Siener. However, the issue with vibe coding or using AI to build new application is that when something goes awry, developers don’t know how to fix it.

“It’s really fascinating to hear all these companies who are now … saying it didn’t make sense for us to rebuild or port this application over,” Siener said. “We can rebuild it now. We can vibe code something to get in production. The problem is they have no context on how it actually works. They didn’t actually write it.”

Honeycomb’s observability tools give users insight into what their AI models or agents are doing and identify where the problem is coming from. The better businesses understand their data, the faster they can address potential issues, Siener suggests.

“Duolingo is a big customer of ours,” he said. “They talk a lot about what it looks like to debug those chat experiences. Having the visibility to understand, here’s the prompt, here’s the sequence of events that happen [and] here are the tokens really helps them troubleshoot faster.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the KubeCon + CloudNativeCon NA event:

(* Disclosure: Honeycomb.io sponsored this segment of theCUBE. Neither Honeycomb nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE