AI Coding Is as Good as You Are: As I experienced
Over the past several months I have used many AI coding assistant tools including Amazon Whisper, GitHub Co-pilot and Codeium windsurf. One thing I can say for sure — I can’t imagine going back without the AI assistance. BUT — it’s not a silver bullet to solve all your coding needs rather it’s a CO-PILOT — you are still in charge.
AI as a Co-Pilot, Not an Autopilot
AI Coding assistant is as good as you are — your expertise on the subject — meaning your ability to understand the generated code, judge the quality of the code, guide AI to fix erroneous code matters and directly influence the outcome you are going to get out of AI.
I am constantly hearing people talk about how software engineering is dead — how programmers will be replaced by AI and Agents. Well, I will be naive to say AI would not replace any software engineering jobs — I think it would. In my opinion, it will replace many software engineering tasks, for example writing the unit test-cases, writing code documentation etc. If that’s the only task you do, then YES, you may be replaced by AI — it’s time to level up and stay in the race. But if you aren’t, AI will make you N times more productive and will make you more valuable software engineer than ever.
AI Coding Assistants Speed Up the Process, But Don’t Think for You
Ghost texts, auto completions — help to write code faster — a lot faster than typing everything — it knows your intent based on the method name and signature and the context in the file. Again, here is the catch — if you use irrelevant or random method names and arguments — AI could do only that much.
AI is capable of generating complex code if you can instruct AI. For example if you clearly design a wireframe and instruct AI what to generate, AI can give you a pretty good head start — like in this example, AI translated my design to Slack block UI.
wireframe
AI generated code to build this modal
This is crazy right? How AI could generate the code for you but imagine, you don’t know how slack block ui works — you won’t be able to instruct as effectively as someone who knows — as a result someone with experience on the subject will get a lot more done, a lot faster than someone who does not.
The prompt I used
Now, on the other hand, take a look at this example where I asked AI to generate Typescript type object from a json data — it speeds up your coding and does not need too much of expertise — your experience does not influence the outcome too much.
Where AI Excels
As I was explaining at the beginning — AI really excels in
- Boilerplate & Patterns: Scaffolding new projects using a framework, generating CRUD operations at once etc. For example, AI helped me create entire landing page.
- Unit Tests & Documentation: AI can help you generate test cases and documentation and speed up your coding. If you are following test driven development — you can use AI to generate test cases based on behaviors.
- Refactoring & Debugging Suggestions: I personally found AI very helpful in code refactoring — for example migrating callback patterns to async-await for entire file all at once.
The Catch
Every now and then you will end up with a situation when you are going back and forth with AI to spit out a code that runs — unfortunately you realize after several conversations you are going in circles — I call it prompt hell
Prompt Hell
As an experienced programmer, you should know when to step away and do your own research to write the code by yourself instead of solely relying on AI for everything.
Conclusion
AI doesn’t replace experience — it amplifies it. The more skilled you are, the more valuable AI becomes as a tool. If you’re a software engineer worried about AI taking over jobs, my advice is to embrace the technology and leverage it to your advantage.
About Me:
I’ve been a software engineer for over 20 years.
Like to share your thoughts? You can find me here:
https://www.linkedin.com/in/connectmithundas/
Co-Creator and Co-Founder of Officebot
