I'm Romain, nice to meet you! on rlibaert

Agentic coding

Sun, 10 May 2026 13:12:54 +0000

AI has definitely transformed a lot of industries and software development is not spared. While this technology is still young and changing, some standards are now emerging.

Agentic code editors

Most IDEs now make use of AI models, either directly or via extensions, to provide powerful features and enhance the development experience. VSCode has been the standard for a long time and is still extremely popular, but the rise of AI helped changing the landscape. Cursor is one of those, but to me Zed is the most innovative (several IDEs are VScode forks, this one is built from the ground up).

Containerized development on Fedora Silverblue

Sat, 15 Nov 2025 21:42:54 +0000

After some time developing containerized applications, I made a habit of trying to run everything in containers. The ability to sandbox applications allows me to keep control over my system and makes me confident to try and discover new software. In this post I show the tools I use to create disposable development environments and to keep my system tidy.

Development Containers

Quoting the website, Development Containers “allows you to use a container as a full-featured development environment”, bringing the benefits of containerization to software development. Basically you can start using a development container by adding to your project a .devcontainer.json file as simple as:

Resume

Sun, 29 Jun 2025 17:25:54 +0000

I have a decade of distilling software problems to their essence and offering simple & efficient solutions.

Skills

Programming: Golang, C/C++, VHDL, Git, GNU Make, Dockerfile, Design Patterns, Testing, Linting
Containerization: Docker, Podman, Nomad, Kubernetes, Dev Containers
Observability: Prometheus, VictoriaMetrics, Grafana, Loki
Databases: PostgreSQL, MariaDB, MongoDB, Redis, Timescale
CI / CD: GitLab CI, GitHub Actions

Experience

Senior Back-End Engineer • RadioKing (NRJ Group)

Since May 2022 • Remote, France

Developed and maintained Golang REST APIs running in a self-hosted Kubernetes cluster
Implemented efficient audio processors in Go using FFmpeg’s libraries & Cgo
Designed a custom Icecast-compatible ingestion network enhancing observability of critical events as well as security & reliability
Architectured an Icecast-based CDN serving hundreds of webradios to 100k+ concurrent listeners
Made audience tracking more reliable by processing asynchronously sessions metadata with FluentBit, providing a resilient & accurate source of insights to advertisers
Provided relevant Prometheus metrics and Grafana dashboards, allowing our team to efficiently investigate issues in the stack and locate root causes
Improved collaboration & progression on projects by being a source of proposals and promoting usage of recognized software design workflows (Domain-Driven Design, EventStorming)
Mentored the team on various tools (Git, Docker, DevContainers) and development best practices
Reduced build time by 20% and artifacts size by 50% on GitLab CI

System Engineer • Stormshield

September 2019 - April 2022 • Villeneuve d’Ascq, France

Advanced flag parsing in standard Go

Sat, 22 Mar 2025 21:42:54 +0000

Over time, a plethora of libraries were developed to build complex CLI applications; spf13/cobra and urfave/cli being probably the most used ones (both have tens of thousands stars on GitHub). While they offer lots of features out of the box, I often found them complex and overkill for my needs.

In this post, we will have an overview of how Go’s flag works and see some advanced uses of it. We will then try to implement parsing of complex values such as comma-separated arguments or slices.

Making sync package type strict using Go generics

Thu, 20 Mar 2025 17:20:54 +0000

The sync package from Go’s standard library provides primitives (Mutex, WaitGroup, …) to help solving synchronization problems. This is an older package and some of the implementations could be different if the generic features were available then.

Indeed, the sync.Pool and sync.Map facilitates concurrent usage for storing and retrieving data in different situations. Those features are pretty elementary and their functionality highly relies on the fact that they can be used with any type, which comes with a few drawbacks. Giving up the type strictness prevents the compiler to detect potential errors or requires type assertion code. The former implies extra attention when writing code and the latter reduces its simplicity and readability.

Must pattern in Go and a generics-based generalization

Tue, 18 Mar 2025 17:20:54 +0000

There is a handy code pattern commonly found in Go codebases, known as Must pattern. To put it simply, for a given Do function returning at least an error, the MustDo function then omits returning the error, panicking instead. For instance:

// package regexp
func Compile(expr string) (*Regexp, error)
func MustCompile(expr string) *Regexp

// package netip
func ParseAddr(s string) (Addr, error)
func MustParseAddr(s string) Addr

In contexts where errors are unlikely to happen, it is a nice way to still handle them while uncluttering the code. Of course this implies the application of the same good practices as when using panic. Those functions are often used to create constant-like objects, such as regular expressions: