Newsletter 74 – 04/2026

Featured

How we hire AI-native engineers now: our criteria – With the engineering work changing so much with the advent of AI, probably we’ll need to change how we hire too, to reflect the new set of tasks expected from most software engineers.

Agents as scaffolding for recurring tasks. – Agents are a beautiful tool to automate recurring tasks while you don’t have time to do a proper tooling to them. This post explains how you can start doing it.

Don’t Let AI Write For You – Many posts around there are being written 100% by AI, this post argues to not do it, since the goal of write can be understand more about problems and build trust.

Eight years of wanting, three months of building with AI – With AI many projects that you dreamed about for many years can now become reality in short time. AI is a fabulous tool to make things real.

MISC

How GitHub uses eBPF to improve deployment safety – Amazing case of how GitHub is using eBPF for o prevent cyclic dependencies.

Let’s add a conditional expression to Go language – Have you wondered how difficult is to add a new expression to a modern language? Of course this can vary depending of each language’s codebase, but this post gives a clear vision of how to do it on Golang and also learn more about how modern compilers work.

Do Humans Still Read Logs? – Interesting case about how SRE Agents barely read traditional telemetry data (Like logs).

Entangled Observability – Through the AI Wormhole – Observability field is changing fast and due to AI maybe everything will be different in a short time. This post gives more than 10 open thoughts about the future of this field.

Containers Are Not Magic: Namespaces From Scratch – Do it by yourself projects are a excellent way of learning how things works, this post explains how you can create your container runtime from scratch and understand more about how it works.

Understanding the Go Runtime: The Network Poller – A deep dive into how network poller works in Go to make block calls feel so smooth to do.

Structured-Prompt-Driven Development (SPDD) – Many people are proposing techniques to manage AI development on large codebases, SPDD is one of them and this post describes the basic concept of it: Prompts as first-class citizens.