This is the story of how Go became the lynchpin for meeting GitHub Copilot’s latency requirements. In this talk I’ll discuss the unique aspects of proxy’ing LLM requests at a global scale, how Go was the ideal fit to unlock low latency, high bandwidth, network connectivity, and I’ll explain how you can take advantage of HTTP/2 in your own applications, including the surprisingly less than intuitive configuration required to do so.
Keynote
That Machine Always Lies: Truth and Fiction in the Age of Artificial Intelligence
Recent events in machine learning are outpacing most people’s ability to keep up with the news, let alone try to analyze the impact.
There actually is a massive shift underway that it is important for everyone to try to understand. However, it is hard to filter out both the hype and the hysteria.
“That Machine Always Lies: Truth and Fiction in the Age of Artificial Intelligence” tries to help explain how we got here, what is going on right now, and what may happen next.
And most importantly, help you and your company figure about what to do about it.
AI is reshaping how software is written, reviewed, and operated. This panel brings together industry leaders to discuss how development practices are evolving, what skills matter most, and how teams are adapting to an AI-augmented future.
For our small backend team managing over 60 Go microservices, AI has become a core part of our development workflow. Recently, I’ve been reflecting on my own journey with these tools and how my colleagues have approached theirs. While everyone uses AI in their own way, many of our experiences overlap in interesting ways. In this talk, we’ll discuss the common themes that have emerged and how individual experiences shape the way we use AI.
Build Your Own USB HID Keyboard with TinyGo (Live Coding)
Can you build a USB keyboard in 5 minutes with Go? Yes — and I'll do it live.
In this talk, I'll show how TinyGo lets you run Go on microcontrollers and turn your code into a real USB HID keyboard. It's the same approach we use in our TinyGo Keeb workshops in Japan, where soldering meets Go.
Green Tea GC: The Insight Behind Go's New Garbage Collector
Green Tea's insight is elegant: scan spans, not objects. But how do you actually implement that? What data structures track which objects are marked vs. scanned? Why does the ownership protocol use three states? And how did span-based scanning unlock SIMD optimizations that were impossible before?
Running AI models locally means no API costs, no data leaving your machine, and no vendor lock-in — but integrating local inference into Go applications has traditionally been painful. In this talk, Bill will introduce Kronk, a Go SDK that lets you embed local model inference directly into your applications with full GPU acceleration — no CGO required. Whether it's chat, vision, audio, embeddings, or tool calling, Kronk gives you the same power as a model server without needing one. To prove it, Bill built a Model Server entirely on top of the SDK, complete with caching, batch processing, and agent support. You'll see live demos from writing your first chat app to driving a coding agent with a local model.
AI coding agents often write outdated Go, generating patterns like manual loops to search slices instead of using slices.Contains. This talk explains why it happens, what "modern Go" means today, and how guidelines and benchmarks can teach agents to generate idiomatic Go code.
Your Tests Pass, But Do They Work? Mutation Testing in Go
100% coverage doesn't mean your tests catch bugs - a test with no assertions still counts. Mutation testing intentionally breaks your code to verify tests detect the change. I'll show how Go's AST and overlay system enable parallel mutation testing without modifying source files.
Lightning Talks
Copied!
A rapid-fire session of short community talks. 5 minutes each.
