It is not a secret anymore. Most developers use AI tools now. If you are not using something like GitHub Copilot, Claude Code, OpenAI Codex, or even just pasting problems into ChatGPT, you are probably in the minority. The stigma has evaporated. Nobody is pretending they wrote every line by hand anymore.
Using AI to write code is just what we do now, like using Stack Overflow was ten years ago except the answers are usually better and you do not have to scroll past three people arguing about whether the question is a duplicate.
This makes sense. The tools are genuinely useful. They save time. They handle boilerplate. They remember syntax I have forgotten for languages I touch once a year. Fighting this feels like refusing to use Google because real programmers memorise documentation. Pointless machismo.
But here is the thing. AI has not replaced developers. Not even close. And I do not think it will anytime soon, despite what the breathless LinkedIn posts from people who have never shipped software would have you believe.
The tools are good, not magic
AI models have gotten remarkably capable. Claude Opus 4.5 can hold context across a massive codebase and reason through problems that would have seemed like science fiction two years ago. GPT-5.2 is genuinely impressive. Copilot autocompletes entire functions that actually work. Codex agents can run tests, fix errors, and iterate autonomously. It is impressive. I use these tools every day and they make me faster.
They also hallucinate. Constantly. Confidently. They will invent APIs that do not exist, reference documentation from some parallel universe, and generate code that looks plausible until you actually run it.
The confidence is the dangerous part. A junior developer might not notice the hallucination. An experienced one spots it immediately because they know what correct looks like.
Design and UX? Forget it. AI can generate a component, but it cannot tell you whether it belongs on the page. It cannot feel that the spacing is off or that the flow confuses users. It has no taste.
It optimises for patterns it has seen, not for the specific humans who will use your specific product. You still need someone who understands the problem, the users, and the context to make those calls.
Power tools, not replacement workers
I think about AI like power tools. When nail guns and circular saws became affordable, they did not make tradies irrelevant. The building industry did not collapse. Carpenters did not all lose their jobs to machines.
What happened was that tradies got more efficient. They could frame a house faster. They could take on more work. The skill was still required. You still needed someone who knew where the nails should go.
A nail gun in the hands of someone who does not know what they are doing is just a fast way to make expensive mistakes. Same with AI code generation.
Copilot will happily autocomplete a security vulnerability. Claude will generate an elegant solution to the wrong problem if you prompt it badly. ChatGPT will confidently explain why your broken code is correct. The tools amplify whatever you bring to them. Skill gets amplified. Ignorance gets amplified too.
Experienced developers use AI as a force multiplier. They know what to ask for. They know how to validate the output. They catch the hallucinations and fix the subtle bugs. They use AI to skip the boring parts so they can focus on the interesting parts.
That is the sweet spot. Human in the loop. AI doing the grunt work, human doing the thinking.
The tradie parallel runs deep
Power tools made tradies more efficient, but you still cannot get a reliable one out to quote on time. The technology changed. The fundamentals did not.
Good tradies are still hard to find. They still juggle too many jobs. They still disappear for three weeks after starting your bathroom renovation. (If you are a tradie reading this and you actually show up when you say you will, I am not talking about you. You are a rare gem. Never change.)
Software is the same. AI makes developers more productive, but the bottleneck was never typing speed. It was understanding requirements, making tradeoffs, communicating with stakeholders, debugging the weird edge case that only happens in production on the second Tuesday of months with an R in them.
AI does not fix any of that. It just means you spend less time writing boilerplate and more time on the hard stuff.
The developers who thrive with AI are the ones who were already good. They use the tools to punch above their weight. They ship faster because the tedious parts take less time. They experiment more because the cost of trying something is lower.
The developers who struggle are the ones who never learned the fundamentals. They copy paste AI output without understanding it. They cannot debug when things go wrong because they do not know how it was supposed to work in the first place.
AI does not teach you to think. It just does some of the typing.
Human in the loop is the point
The phrase human in the loop gets thrown around in AI safety discussions, but it applies perfectly to development. You want a human making the decisions. You want AI handling the execution. The human understands the goal. The AI generates options. The human picks the right one and catches the mistakes.
This is not a temporary state while we wait for AI to get smarter. This is the correct architecture. Software exists to solve human problems. Humans need to stay involved in defining those problems and validating the solutions.
Fully autonomous code generation sounds cool until you realise nobody is checking whether the output actually does what users need.
The best AI assisted workflow I have found is simple. I describe what I want. AI generates a first pass. I review it, fix the issues, and ask for refinements. We iterate until it is right.
The AI does in minutes what would have taken me an hour of typing. I do in seconds what the AI cannot do at all, which is decide if the result is actually good.
That is the future. Not AI replacing developers. Developers with AI replacing developers without it. Same as power tools. Same as Stack Overflow. Same as every other productivity improvement in the history of the industry.
The humans are not going anywhere. We just have better tools now.