Rygen Technologies is a premier supply chain technology provider, that enables any company to quickly, easily, and efficiently execute and manage freight, connect with partners, and unlock data insights for their supply chain. With emphasis on advanced technology and exceptional customer service, Rygen is fueling the creation of smarter, data-driven solutions that deliver real value and solve real business problems.
Hot Take: we still need junior developers...
I don't know of another time in recent history where a technology emerged that not only disrupted one industry, but literally every industry in some way. AI has certainly done that. In fact, it's almost annoying that everything we see online today is either made by AI, is talking about AI, or is about a new AI startup. But anyway, here I am talking about AI....
Someone recently asked me about what percentage of code was AI generated at Rygen. This is a logical question to ask based on all the trends we see in the industry. We all know that this is massively intriguing for management - what a great way to reduce labor costs, right? Agentic swarm coding for the win, right!? Wrong... kinda.
Back to my original question, if I said it's less than 10% of our code is AI generated, is that bad? Or if I said 80% of our code was AI generated, is that good? The reality is, for some engineers, it might be less than 10% and for others, it might be greater than 80%. Which engineer is doing it right?
I guess let's just throw out the obvious point here. AI generated code is a massive efficiency multiplier WHEN USED RIGHT. But also, AI generated code introduces a massive amount of technical debt WHEN USED POORLY. So, then what does it mean to use it right?
For an engineer to vet the quality of AI generated code, that engineer needs to be experienced enough to know what the correct answer should look like. This is the only way to validate that what is generated is quality, maintainable, and scalable code. If that engineer isn't experienced or knowledgeable enough, then the code is seen with ignorance and more than likely, that code will be copy and pasted into the codebase without a second thought.
In 2025, a study by METR looked at the efficiency gains of engineers that used AI tooling vs those that didn't. The findings were pretty interesting in that they showed that while the experienced engineers using AI believed they were 20% faster, in reality they were almost 20% slower. WHAT THE HELL? A classic productivity paradox. To me, the answer is obvious. The engineers were taking the time to craft proper prompts and instructions, asking for refactors, code reviewing, and vetting the output. Essentially, they were paying off the tech-debt that was accumulating right away.
What I am implying here is that to really use AI effectively in the enterprise world, AI absolutely requires highly skilled engineers to vet and validate outputs. I'm not talking about creating MVPs, prototypes or vibe-coded million-dollar app ideas. I'm talking about scalable applications meant to be built upon for decades to come.
So then this begs the question, how do we get highly skilled and knowledgeable engineers in the future? Well, we hire junior developers. We train them. We coach them. We teach them responsible AI usage and give them guidelines to follow. We shouldn't let AI be the reason CS grads end up working at Chipotle.
There is still something to be said about learning to program the old-fashioned way. I spent so many nights bashing my head against my desk, trying to figure out a seam-carving algorithm for a computational photography course in grad school. Something that AI probably could have done in 2 seconds for me. But it gave me the skills necessary to know what good code should actually look like. Knowing what the correct answer is and should look like is key. One of my favorite examples of this lies in the question: what is the difference between a junior and senior software engineer? The senior software engineer knows "what" to Google.
Let's keep the software engineer career path going strong. Let's raise up juniors to become principals. Let's use AI safely and effectively. Engineers + AI > AI.
Article by: Tony Winters, CTO at Rygen
Sources:
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
https://www.nytimes.com/2025/08/10/technology/coding-ai-jobs-students.html
.png)