AI in software development - realistically, it can be a force multiplier, but not a replacement

I'm a software engineer and architect for 24y. Was actively using AI in my development work last few weeks, mostly experimentation. I wanted to share my honest experience here as many are continuously arguing about how awesome/bad it is. There's been a lot of doom and gloom about AI **replacing** developers... specially on social media. Mostly lies. Here's the thing - AI cannot replace mid/above developers, not yet. It frequently forgets the context, even for a tiny application with 3 classes. You can remind it, using lots of tokens (more money), but almost every time the output is going to be different... and needs massive amount of fixing - which mostly an experienced person can handle properly. Now, after you take care of that, you're not gonna let AI touch it again... you get the point. What it can do however, is become a good(not great, yet) force multiplier. In my experience, it can make a small team more productive, especially in the initial stages of development. Say, a 2-person senior dev team using AI effectively can now handle workloads that previously required 4-5 people (typically 2 senior + 2-3 junior/mid-level devs). Generally, don't expect it to do the following (I would grade them 0-3 out of 10, currently) : * Understand complex business requirements * Design system architecture * Debug tricky issues * Make critical technical decisions * Ensure code quality and maintainability Bottom line: It's making some changes into how we work... but it's more like having autocomplete on steroids rather than a replacement for developers. It can make good/sr. developers more efficient, but cannot make them obsolete. Not yet.