The network is unpredictable. Your response shouldn't be #RTC #WebRTC

3 Apr 2026 · 1 min

Watch on YouTube

1:07

In this brief talk, the speaker argues that relying on large language models or AI to solve WebRTC problems is ineffective, emphasizing that the network is inherently unpredictable and cannot be fully controlled. Instead, the speaker highlights the importance of visibility into network behavior and presents RTCStats as a lightweight, observability-focused tool that extracts raw getRTC data, enabling engineers to pinpoint the root causes of call failures without imposing a heavy platform. The talk concludes with an invitation for viewers to share their most frustrating "ghost" bugs in WebRTC applications.

Chapters

In the WebRTC world, there is a lot of noise. There is no silver bullet here, and the promise of brute-forcing it with LLM and AI as some miracle fairy dust isn't really working. I know because we've tried. Here's the truth. The network is a variable you cannot always control. What you can control is your visibility into it. We built RTCStats for the engineers who are tired of running after metrics. We don't try to sit in the middle of the media or promise impossible fixes. Instead, we act as a quiet observer, extracting the raw getRTC data you need to actually solve problems. Our goal isn't to be a heavy-pushy platform. It's to give you the precision tools to identify why a call failed and fix the root cause. No fluff, just real diagnostics. Here's a question for you. What is the most frustrating ghost bug you've had to chase in your WebRTC application?