... it's the end of 2025. And the development world as we know it is at turning point. It's been at turning point a lot of times since I've started doing this some 16 years ago, and a lot more times before that. But this one time is special, this one time is different: AI threatens to replace us.
Relax, I'm dramatic on purpose, we're not going anywhere. I'm writing this post mostly for younger folks who I don't envy because they're in a position where they are threatened by, scared by, and forced into the AI bubble. People are losing their minds over whether AI will make them obsolete, they're listening to false prophets who are telling them that they should be learning a real craft, a physical skill that won't be touched by AI any time soon. But what's the reality? Reality is that AI is a tool (last time I checked). A powerful tool and it could be a great asset to a developer who knows what they're doing. If you don't know about the basics, security, user experience, performance... no amount of AI will help you make a good app. That's the truth, and anyone who's telling you otherwise is likely fearmongering to raise their importance or to sell you an AI-powered service or one of their dime-a-dozen courses. These people are not your friends!
I'm not saying you need to avoid AI until it disappears. On the contrary, it's here to stay - just maybe not in the areas AI doom merchants want you to believe it will. You should definitely learn how to work with AI. Large Language Models (LLMs) changed the game for me: I can get to the prototype faster, I can debug faster. I abstracted away boring multi-file edits, tests, and boilerplate, which are generally time consuming. Do I work more? No. Do I output more? Yes, but probably not as much as my CEO would like me to. I use it to find myself the time to learn core stuff that I'm missing, stuff that AI maybe knows but is unsure how to apply it (or apply correctly). I'm improving my infrastructure and backend knowledge, I use AI to brainstorm and then have it explain the choices it made. Then I question its choices, which are often either overkill or outright wrong. And this happens more times than any of the doom merchants would like to admit.
It's easy to pay $200+ a month and have Claude or any agent write the app for you, but what happens when you get a data breach and you don't understand the code well enough to fix it? When you can't explain to your users what went wrong because the AI made decisions you never questioned? Will you be able to sell your service when everyone can use that same $200+ subscription to build their own? Is that sustainable?
I don't think so.
Instead, people will still need well-thought-out software that they can use without a hassle of setting up 10 cloud services, or worrying about backups, availability, or disaster recovery.
Will our lives change?
You bet they will. Which way - that's up to you. You need to decide if you'll chase the agent-of-the-day or you'll use the AI to produce boring solutions that actually work.
Your call.
I got this figured out for myself. I think I'll have a lifetime of work fixing vibe-coded messes false prophets will inevitably create. It'll likely be boring but, sometimes, boring is good.
Top comments (0)