on Kubernetes, Finops, Cloud, Devops, Cost optimization
There’s a dirty secret about Kubernetes that nobody talks about at KubeCon: most clusters are wildly over-provisioned, and most engineering teams have no idea what their workloads actually cost.
on Ai, Mcp, Llm, Integration, Protocol
If you’ve spent any time building AI-powered applications in the last year, you’ve probably run into the same problem dozens of times: every tool, every data source, every external service needs its own custom integration. Your AI assistant needs to talk to your database? Write an adapter. It needs to call your APIs? Write another one. Search your files? Yet another bespoke connector.
on Security, Zero trust, Cloud, Devsecops, Identity
The old security model was built around a perimeter: inside the firewall, you’re trusted; outside, you’re not. For decades, this worked well enough. Corporate networks were physical places. Servers sat in data centers you controlled. Employees came to the office.
on Webassembly, Wasm, Cloud, Edge computing, Performance
WebAssembly was introduced to the world as a faster way to run code in browsers. C++ games, video editors, image processing — things that JavaScript couldn’t handle performantly. That was the pitch in 2017.
on Rust, Performance, Systems programming, Go, Python
I’ve spent the last three years migrating production services from Python and Go to Rust. Some migrations were triumphant. A few were disasters we quietly rolled back. Most were somewhere in between — significant wins that came with real costs.
메모는 기억보다 확실하다