In 5 years We’ll Code Everything with GenAI

I recently completed a code assessment using Android native views, where I worked with a JSON adapter to implement a scroll list with RecyclerView. 

I decided to use Copilot to help me out, but I still need to review everything carefully. Honestly, working with Copilot feels like coding with my younger self.  

Eager to get things done but need guidance. 

The notion that LLMs can’t produce reliable code is short-sighted.

 Let’s face it—most junior and mid-level developers don’t either.

That’s why software engineering has spent decades refining automated and semi-automated tools to ensure code is production-ready.

💡While LLMs might be unreliable as software, it feels more like working with an intern who’s eager to get things done but needs guidance. 

Generative AI has seen huge investments from tech companies aiming for mass adoption. Despite the slow returns, failures, and unreliability of LLMs.

But, I believe that those who learn to integrate LLMs effectively into their workflows will reap significant benefits.

In the past few months, most of my "programming" has shifted to writing in English—prompting, then reviewing and editing the generated diffs.

 I’m also doing a bit of "half-coding," where I write some of the code, drop a few comments so the LLM has context. Then I just tab through the completions. Sometimes, I get a 100-line diff that nails it, something that might have taken me over 10 minutes to write on my own.

As we look toward a future where AI churns out code at a breakneck pace.  

The future of software development jobs lies in codebase comprehension and system maintenance. It's not about who can churn out code the fastest, but who can navigate the codebase.

💡To truly understanding a codebase means going beyond syntax.

It’s about grasping the underlying logic, architecture, and design patterns that hold everything together.

Remember that coding serves as a means to communicate with computers.

While AI may write the initial code, it is human developers who will be tasked to maintain and scale the codebases of tomorrow.

Thanks for reading 🖖

Come say hi 👋 I love to chat at LinkedIn