The State of Software Engineering in 2026

Bro has fallen so far. He used to be a geeky kid in his room writing python for fun. Now he lives in Dubai and sells AI slop on the internet That’s depressing af

"Productive" does not mean writing as much code as possible, it means writing systems that are robust and need little maintenance and refactoring.

I've been coding for 18 years. Worked in financial, government, insurance, you know name it. Needless to say coding is no longer what it used to be. AND I still love it. Here are my top recommendations for new devs - Write every new concept you're learning by hand for the first time. - Learn principles, not just hows, but whys. Not how do I implement an interface, but why am I implementing an interface? Does it actually need to be an interface? - Question everything until you understand it to it's core. Old school coding, you had to deal with bugs that forced you to look into depth you never thought you would and in the process you would learn so much. Now, if the bug is fixed with a prompt and you're not sure why, your fundamentals are weak. - Write 1 project by hand. Make it GOOD. like make a GIT Repo, from scratch. You will learn SO MUCH. After that, AI can't write code that will easily confuse you. - Don't accept code from AI that you don't fully understand. And question AI on it's decisions. Do not assume it's more logical than you.

Don’t use AI especially when you are beginner, hear me out. I am not a luddite, AI just another tool but AI can give you the illusion of competence. Let’s say you ask AI to build react app and it spew the code and you are feeling good because you just “create” react app but did you? did you really know react inside and out? best way to familiarize yourself is write it yourself, learn to read documentation, organizimg your knowledge, how to debug, how to make your code readable and consistent. This skill is more valuable than knowing hiw to use most recent AI agent. The head of claude code doesn’t write any code yes but he is a senior engineer, he know inside and out his system. Obviously the same thing doesn’t apply to you. How can you supervise if you don’t even know your stuff? AI tool is just another tool, you’ll need to learn it at some point but don’t get trapped in hype train and completely ignore your fundamentals. Writing code is the easist part of the job, maintainimg your code is the hardest. Don’t be an engineer who are useless when Claude is down. Don’t get dependent to AI , use your natural intelligence

I’m working on a migration project to Python that was farmed out to Cap Gemini. They brought in Jr. devs from India using LLMs. Their code was full of deprecated patterns, security issues and was generally an untestable mess but it produced the same results as the original code. I genuinely had to ask if they knew python and they admitted they didn’t. Management would have accepted this and not known if they didn’t already have someone on staff who could audit this. As soon as everyone is reliant on LLMs the big providers will rug pull everyone, it’s already started with their pricing model changes. This is the path we’re on now.

I don't think I want to be a plane where the autopilot code was vibe coded.

@1min “we're no longer really responsible for writing….”, as a contractor in investment banking, lawyers may find me responsible for any line that loses the bank money.

Request: A 2026 Software Dev Roadmap. The era of AI has flipped the script on the traditional 'HTML/CSS/JS' progression. How should we be learning system design and coding fundamentals today? We need your insight!

It was nice listening to your thoughts - thanks for sharing them Given that we are shifting from writing code to reviewing code, what pro tips do you have doing thorough code reviews using coding agents?

Congratulations on 2M subscribers!