Posts

Last week, I attended the QCon AI 2026 conference in Boston. One theme kept coming back throughout the conference: what happens to engineering organizations when code generation becomes cheaper, but review, evaluation, infrastructure, and ownership remain expensive? This question came up in different forms throughout the conference. Sometimes it appeared in discussions about code review. Sometimes it appeared in talks about agent evaluation, infrastructure, or the changing role of software engineers. But to me, many of the conversations pointed in the same direction: AI is not only making software development faster. It is also shifting the bottlenecks. ...

Struct initialization in Go is simple, but enforcing mandatory fields can quickly become tricky in larger codebases. Let’s walk through the problem and explore a possible solution. Named field initialization One common way to create struct instances is by specifying fields by name. type Vertex struct { X int Y int } fmt.Println(Vertex{X: 1, Y: 2}) // {1 2} This allows us to omit fields at creation time, which will make Go fall back to the default value for the omitted fields: ...

I finished reading the “AI Engineering” book by Chip Huyen some time ago and it was quite a long read for a technical book. I’m interested in AI technology and this book received quite a lot of hype which meant I had to read it. Expectations & Reality At the beginning, I was a bit skeptical of the idea of a book about AI engineering now - an area that evolves very rapidly and which appeared just a few years ago (referring not to AI engineering in a general sense, but to the surge of interest following the release of ChatGPT). I expect that many things that are considered standard and best practices today may change a lot in the near future and that would make the book obsolete fast. But I was surprised by how Chip approached writing on this topic: the book seriously tries to talk about fundamentals of the technology and it often talks about how certain aspects of it should be reasoned about instead of providing step-by-step guidance. Maybe not everything in this book is like this: some model architectures or concepts like “LLM as a judge” might disappear in the future. But overall, the book has the potential of remaining relevant. ...

This is a story of some relatively big code refactoring and my experience of using Claude code for it. At Uber, we have a Go service for generating financial statements for users (let’s call it service X). Over time, this service grew into a pretty hefty monolith service which includes data fetching, massaging and presentation logic. We had to build a new service (service Y) for a different type of reporting that should re-use the data fetching part of the X service. Unfortunately, this part had typical features of legacy code - written long time ago, not well maintained but at the same time was updated with some frequency which made the quality rather worse than better with every diff. ...

A few weeks ago, I watched “Deep Dive into LLMs like ChatGPT” by Andrej Karpathy. I found it EXTREMELY valuable. It’s the most concise material I’ve seen so far on this topic. If you have some knowledge of software and want to learn about LLMs, this video is the one to watch. It explains all the main concepts and fundamentals of this new tech. ...

Hello, Internet This is the first post on this blog just to test how it looks like. This is Dolomites: