Learn how to build AI Agents with @guiferreira.me at NDC Copenhagen 2026. The Early Bird offer ends this Friday, 27 Feb! Tickets -> ndccopenhagen.com

The 80/20 Rule (Pareto Principle) is the secret weapon of the world's most efficient dev teams. Here's why you're likely wasting 80% of your energy.

System design for Agentic AI. A helpdesk triage agent that retrieves facts, calls tools, and knows when to hand off.

80% of your results come from 20% of your effort. Stop wasting 80% of your time on the features nobody uses. Subtraction is the ultimate productivity hack.

The biggest mistake in tech: Starting with the "How." Dev: "I want to use Rust/Microservices/AI." Lead: "Why?" Dev: "Uh... because it scales?" (This is retrofitting).

Focus is the ultimate force multiplier in IT. 50 engineers with focus will outperform 5,000 engineers without it every single time.

In 2014, WhatsApp had 900 million users. How many engineers do you think it took to handle that scale? 5,000? 1,000? Not even close. It was 50.

System design for Agentic AI. A helpdesk triage agent that retrieves facts, calls tools, and knows when to hand off.

"What part of 'one click' did we not understand?" - Jeff Bezos. If your engineers are adding "confirmation steps" to your vision, they aren't helping. They're complicating. Hold the line.

Statistics that should worry everyone in IT: 30% of software developers have a diagnosed mental disorder.

Software "Sprints" are a lie. You can't sprint every day for 5 years. That's just a recipe for a system (and a team) that eventually attacks itself.

Most developers are addicted to complexity. We applaud "clever" systems and put them on our CVs. But complexity is a liability, not an achievement. Let's talk about why we get it wrong.

An AI agent is not a chat response. It is a loop. Load context, reason, execute tools, update memory, repeat. Once you design the loop, agentic AI becomes just system design again.

AI agents aren't just chatbots. They're loops. Here's the agentic loop explained 👇

Building an AI agent should not start with prompts. It should start with architecture. Context diagrams, application boundaries, and component design turn "AI magic" into systems your team can actually build and maintain.

Want to build AI agents in your applications? Don’t start with prompts. Start with architecture. Here's why 👇

System design for Agentic AI. A helpdesk triage agent that retrieves facts, calls tools, and knows when to hand off.

30% of devs have a diagnosed mental disorder. We need to stop applauding "complex" architectures that lead to burnout. Simplicity is a mental health requirement.

"Don't Repeat Yourself", it’s not about cutting lines of code. It's about avoiding knowledge duplication. 🧠💡

Happy New Year 🎉 Thank you for being here this year. I wrote a short reflection on 2025. Link below.

Architecture diagrams are one of the fastest ways to think, communicate, and plan. A few boxes and arrows can clarify more than a full page of text. Every developer should master them.

Architecture diagrams aren't just pretty boxes and arrows. They're one of the most underrated tools in software development.

Are you aware your code assistant could be leaking secrets without you noticing? Connection strings, API keys, config files… all can be sent as context to AI providers.

Wishing you a simple, calm Christmas. Thank you for being here this year.

Want Cursor to act less like chat and more like a teammate? Use a two-rule workflow: plan everything, then execute commit-by-commit. More control, better output.

💡 AI tools are powerful, but your secrets aren't safe by default. Audit your project, move sensitive info to secret managers, and configure every code assistant to ignore those files. Stay productive and secure.

Choosing a diagramming tool isn't about which one looks the nicest. It's about matching the tool to your workflow.

30 minutes to solve a 5-minute problem? The reason: not knowing your tools. 🛠️ Learn the language features you already have. Mastery = simplicity + less frustration.

Your code assistant is powerful, but it might be sharing secrets with AI providers. Connection strings, API keys, secrets… all are at risk.

🎥 If Your Integration Tests Feel Impossible to Debug, Watch This! Ever tried debugging an integration test, only to get nothing in your output window? No logs, no console messages, just confusion. That's exactly what xUnit v3 fixes with CaptureOutput. 👀 Watch the full video on YouTube 👇