Blog Posts

Latest articles and insights

Use Claude Opus 4.8 to prioritize review attention without replacing engineering judgment.

Turn Pull Requests into Risk Maps with Claude Opus 4.8

Răzvan Mareș8 min read

Most pull requests are not equally dangerous. A one-line configuration change can expose an internal service. A 400-line refactor can be mechanically safe. A harmless-looking default value can alter ...

GitHub Actionsci-cdTypeScript+2
Practical techniques for turning vague prompts into dependable engineering collaboration.

How to Get Better Results from Claude in Real Engineering Work

Răzvan Mareș8 min read

Claude is often described as a coding assistant, but that framing is too narrow. In a real engineering workflow, it can act as a requirements analyst, debugging partner, test designer, architecture re...

nodejsTypeScriptAI/ML+2
Design .NET agent systems as observable workflows instead of unpredictable chat loops.

Orchestrating AI Workflows with Microsoft Agent Framework 1.0

Răzvan Mareș8 min read

AI agents are often introduced as conversational features: provide a prompt, call a model, invoke a tool, and return an answer. That framing is useful for a demo, but it is too small for a production ...

.NETcsharpAI/ML+2
A practical architecture for parallel agent work, isolated branches, and trustworthy integration.

Orchestrating Multi-Agent Coding Without Turning Git Into a Battlefield

Răzvan Mareș8 min read

Software teams have always parallelized work. One engineer builds an API, another changes the UI, and a third prepares tests or deployment automation. Multi-agent coding extends that familiar pattern,...

GitHub Actionsci-cdgit+2
How I structure compute, networking, storage, and operations in a k3s homelab.

From Proxmox VMs to k3s Workloads: Designing a Practical Home Cloud

Răzvan Mareș8 min read

A homelab can begin as a spare machine under a desk and quietly become something more ambitious: a small, imperfect cloud built from hardware you can hear working. My stack is centered on **Proxmox** ...

traefikproxmoxCloudflare+3
How to give production agents room to act without giving them the keys to everything.

The Permission Envelope: A Control Plane for AI Agent Actions

Răzvan Mareș8 min read

Production AI agents do not fail because they are too autonomous. They fail because their autonomy is undefined. An agent that can read a support ticket, inspect an order, issue a refund, update a CR...

PostgreSQLAI/MLrest-apis+2