One of my readers replied to an email I sent a couple of weeks ago and we got into a brief discussion on what I’ll call, Skills Erosion.
They brought up the point that by leaning too heavily on AI to generate code, people were losing their edge.
It’s a good point that’s top of mind for many devs. I’m guessing you’ve thought about it too. After all, if AI writes all of our code, how are we actually learning anything?
The exchange made me go down a rabbit hole and I found the data quite interesting.
We all know what Vibe Coding is, so I’ll save the explanation.
One of the biggest issues with vibe coding is that it creates the Illusion of Competence.
I mean, you feel 5x more productive with AI on your side, right? (I know I do!)
But reports from Veracode (2026) show that 45% of AI-generated code contains security flaws.
The companies that rely on AI to vibe code their products are shipping code that introduces security events just waiting to make the news. This is what happens when we trust the machine more than our own earned expertise.
It’s no surprise then to hear that some teams and companies are starting to apply the brakes and slow down their AI adoption.
But this is the catch. Not all companies are slowing things down. Some are ramping it up. (Look at Amazon’s announcement last Thursday – US$200bn investment in AI infrastructure in 2026).
Where does this leave us as devs?
I believe a balance needs to be found. And I said as much to our reader.
You can’t just be a hold-out.
At the end of the day, so many of the people holding the keys to our pay cheques expect us to use AI. Hiring Managers, CTOs, CEOs, Shareholders, Investors – you name it.
If you refuse, you look obsolete.
The solution? Be the architect and auditor, not the operator.
The developers who will come out on top are the ones who:
- Spot the hallucination: know why that SQL query is inefficient.
- Question AI recommendations: don’t treat the generated code as gospel. Question design decisions (or a lack thereof).
- Refactor the mess: turn spaghetti code into clean architecture.
- Secure the build: know where the vulnerabilities hide.
- Sharpen the saw: keep your skills sharp outside of AI usage. Keep learning, keep growing.
I firmly believe that the AI hype will plateau. We’re already starting to see the cracks.
The real questions to ask yourself: where will you be when things start shifting back in our favour? Will you be the senior dev ready to jump in and save the day?
Don’t let your skills erode: use the tools but master the craft.
What do you think?
Join me in the community for a chat on the topic. You can also check out a post I created for people to share their thoughts on AI + LLMs + coding.
Julian
This was originally sent to our email list. Join here.