Happy Monday, Craftsmen, time to set the tone for the week. Accountability and moving forward is how we grow - together. Show up and share progress. How to join (it's simple): 1️⃣ Post your 1–3 goals for the week in the comments. 2️⃣ Update your own comment on Friday or Sunday with your progress + biggest win. 🏆 End of month, the Top 3 on the 30-Day Leaderboard win $500 / $350 / $100 prizes. Done or not, progress counts. Post your goals below, then come back at the end of the week to share proof. Let’s build momentum together! Good luck, fellow Crafters. PS: Add a GIF for your Monday mood 👇

GPT-5 launched this week with the usual fanfare and “revolutionary breakthrough” claims. ​ The reality? “Overdue, overhyped and underwhelming” – that’s how AI researcher Gary Marcus described it just hours after launch. ​ THE BRUTAL X REALITY CHECK Within hours of the GPT-5 livestream, prediction markets told the real story. OpenAI’s chances of having the best AI model dropped from 75% to 14% in one hour. Users on X flooded the platform with harsh reactions. They called it a “huge letdown,” “horrible,” and “underwhelming.” GPT-5 even gave wrong answers when asked to count letters in “blueberry.” ​ WHY AI AGENTS WILL NOT REPLACE YOUR DEVOPS WORK, FOR NOW Gary Marcus predicted exactly what we’re seeing. His analysis shows why AI agents pose zero threat to DevOps professionals. The key issue is that current AI systems work by copying patterns, not real understanding. Marcus calls this “mimicry vs. deep understanding.” AI can copy the words people use to complete tasks. But it has “no concept of what it means to delete a database.” This matters for DevOps work. When you debug a networking issue between services, you don’t just run commands. You form ideas about how systems behave under load. ​ An AI might know kubectl get pods syntax. But it doesn’t understand why pod networking fails. It doesn’t grasp what this means for other services in your environment. ​ Marcus notes that complex tasks involve multiple steps. DevOps work has many steps: deploy, monitor, check results, maybe rollback. And this is why AI agents are not going to replace us anytime soon. Large Language Models (LLM's) are relatively simple input-output systems. They are useful, but the problem is that their output is unreliable. So far it is nearly impossible to make an LLM reliably give the same output in the same format, especially when the input can vary wildly. Since DevOps work at its core always has complex tasks with multiple steps, one mistake in the chain could cause a system-wide outage.

This post is for everyone who's still on levels 1 to 3. Go on, say hello :-) Anyone who's level 4 or higher is only allowed to use the reply function. Let's see if this works here -Mischa