AI Coding Assistants: Cheating or the Future?

We need to talk about AI Coding Assistants on the job, and in the job interview 🚩

‍Sore, but true: a month ago a client rejected one of our candidates because they suspected he was using an AI coding tool during the technical assessment. I was taken aback – this felt like rejecting a developer for using Stack Overflow for advice.

The reality is that AI coding assistants have become as commonplace in software development as version control. These tools have moved from novelty to necessity, demonstrably improving both code quality and developer productivity.

For those less familiar: AI coding assistants are tools that integrate directly into your IDE, using large language models to understand context and suggest code completions in real-time. Think of them as extremely sophisticated autocomplete that can write entire functions, suggest optimizations, and even explain code behavior.

The market leaders are clear:

GitHub Copilot leads with enterprise adoption, showing up to 55% productivity boost in internal studies, and reported $400 million in ARR in November 2024  

Cursor.sh combines coding assistance with powerful AI-driven code navigation, and reported $100 million in ARR at the end of 2024

Amazon CodeWhisperer has captured significant AWS-centric teams

The real question isn't whether these tools are effective – they demonstrably are. The question is: how should we approach their use in professional contexts?

On the job, the case is straightforward: AI coding assistants are productivity multipliers. They handle boilerplate, suggest optimizations, and help developers focus on higher-level problem-solving. Just like we don't expect developers to memorize every API, we shouldn't expect them to write every line from scratch.

The interview context is more nuanced. While blanket bans feel shortsighted, we need clear guidelines. My take is that the interview context should be as close as possible to the real-life work context. Surely it's about getting the job done? That said we complement coding tests with evaluation of problem-solving approaches, system design, and code review capabilities – skills that remain distinctly human.

The transition is happening whether we embrace it or not. Forward-thinking engineering teams are already incorporating AI coding assistants into their workflows and updating their evaluation processes accordingly. Those who don't risk missing out on talented developers who've integrated these tools into their workflow.

What's your take? Is it cheating? Or is it an evolution in how coders do their job?

Bonus points if you describe where you see coding 5 years from now!