AI Will Take My Job, But Not Today

Recently, I noticed new feelings and thoughts about AI in Software Development. I felt upset about recent news such as “Programming no longer needed in the nearest time” and “AI will replace programmers”. I took a lot of time to develop a career as a software developer, and now AI will replace me. We live in a fantastic world where AI can solve many technical problems, and this technology makes our lives better. I use it almost every day, but not for production-ready programming. Let’s speak in the facts of why AI wouldn’t replace me in the closest time. Results consistency: AI could provide you with working code day after day, without pauses and hunger, but it does not consistently provide good results. The generated code could be laconic, chaotic, and long, with different naming and logic structures. Maintainability of results: AI could provide code that solves your problem but generates code that is hard to maintain. Software Development has not ended with programming. Developers also maintain code by improving performance, fixing bugs, creating new features, etc. Too long code: Poor maintainability is the side of the problem. AI generates a lot of code with many comments in one file. New functions are added to the end of a file or somewhere. In common cases, when the developer works on functions, they could adopt them to reuse existing methods or extend them, accurately structure files, classes, and methods to have well-organized projects. Hallucinations: This is not a solved problem for LLM models like ChatGPT. AI could invent something that programming languages and libraries don’t have. During my workflow or requirements, AI often suggests methods in libraries that didn’t exist before. If you point that problem to the Assistant — it invents new, not existing methods, and very apologies! Knowledge limitations: AI generates code for the most common scenarios. AI doesn’t know much about new versions of libraries or languages. It could generate only what he learned for most common cases. One good example is Exposed (ORM framework). In new versions of that framework, they changed how data is retrieved. They changed API. Old methods were deprecated for some time, and old methods were removed after some period. So, AI now generates invalid code that doesn’t work. Business domain and company knowledge: AI should know how to solve problems specific to your company. Problems are not solved only by AI: collaboration, research, and expertise should happen to make code that solves problems. AI couldn’t invent ways to help you solve problems effectively, probably without code. It could provide you with an average solution for problems that could be resolved in different, easier ways. At the end AI has limitations. Some problems could be solved with prompt engineering, but others could not; different approaches could be made, like AI agents. I believe in the AI future. It will take my job in its current state. Developer work will be changed insanely from how It was decades ago when modern high-level programming languages were created. The same will be true with the adoption of more powerful AI tools. Do you want to know what matters in a Developer’s work? Read my past stories, and follow me to read upcoming ones.

Apr 14, 2025 - 17:43

Recently, I noticed new feelings and thoughts about AI in Software Development. I felt upset about recent news such as “Programming no longer needed in the nearest time” and “AI will replace programmers”. I took a lot of time to develop a career as a software developer, and now AI will replace me.

We live in a fantastic world where AI can solve many technical problems, and this technology makes our lives better. I use it almost every day, but not for production-ready programming. Let’s speak in the facts of why AI wouldn’t replace me in the closest time.

Results consistency: AI could provide you with working code day after day, without pauses and hunger, but it does not consistently provide good results. The generated code could be laconic, chaotic, and long, with different naming and logic structures.

Maintainability of results: AI could provide code that solves your problem but generates code that is hard to maintain. Software Development has not ended with programming. Developers also maintain code by improving performance, fixing bugs, creating new features, etc.

Too long code: Poor maintainability is the side of the problem. AI generates a lot of code with many comments in one file. New functions are added to the end of a file or somewhere. In common cases, when the developer works on functions, they could adopt them to reuse existing methods or extend them, accurately structure files, classes, and methods to have well-organized projects.

Hallucinations: This is not a solved problem for LLM models like ChatGPT. AI could invent something that programming languages and libraries don’t have. During my workflow or requirements, AI often suggests methods in libraries that didn’t exist before. If you point that problem to the Assistant — it invents new, not existing methods, and very apologies!

Knowledge limitations: AI generates code for the most common scenarios. AI doesn’t know much about new versions of libraries or languages. It could generate only what he learned for most common cases. One good example is Exposed (ORM framework). In new versions of that framework, they changed how data is retrieved. They changed API. Old methods were deprecated for some time, and old methods were removed after some period. So, AI now generates invalid code that doesn’t work.

Business domain and company knowledge: AI should know how to solve problems specific to your company. Problems are not solved only by AI: collaboration, research, and expertise should happen to make code that solves problems. AI couldn’t invent ways to help you solve problems effectively, probably without code. It could provide you with an average solution for problems that could be resolved in different, easier ways.

At the end

AI has limitations. Some problems could be solved with prompt engineering, but others could not; different approaches could be made, like AI agents. I believe in the AI future. It will take my job in its current state. Developer work will be changed insanely from how It was decades ago when modern high-level programming languages were created. The same will be true with the adoption of more powerful AI tools.

Do you want to know what matters in a Developer’s work? Read my past stories, and follow me to read upcoming ones.