For the past two months, I’ve been interviewing candidates for a Senior Software Engineer role. Simply put, our industry is facing a real competence problem.

I’ve met candidates with many years of experience, impressive LinkedIn profiles, active GitHub accounts, and well-known companies on their resumes, but they still can’t answer basic technical questions like:

• Explain how memory management works in programming language X.
• What is dynamic dispatch, and when would you use it?
• Describe the difference between concurrency and parallelism.

These aren’t advanced topics or tricky questions. They are fundamentals you should have mastered long before reaching a senior level.

At this point, it’s hard to ignore the pattern. Many engineers rely on AI to do their thinking for them. Let me be clear: If you can’t solve problems without AI, you are not a senior engineer. You’re just serving as an interface for a language model. This isn’t what the future of work should look like. This is a sign of cognitive decline.

What worries me isn’t only that interviews are getting worse. The bigger problem is what happens when these engineers ship code to production. Software today runs:

• airplanes
• medical devices
• financial systems
• critical infrastructure

When engineers no longer understand the systems they build, failures don’t just cause crashes. They can actually harm people. It might seem like a joke now, but when systems start failing in ways no one can expect because the AI wrote the code, it won’t be funny anymore.

Use AI. Of course. It’s genuinely useful for boilerplate, exploring unfamiliar APIs, drafting documentation, writing tests, or getting unstuck on a tricky problem. I use it myself. The issue isn’t the tool. It’s becoming dependent on it.

But if AI has replaced your ability to do things like:

• explain basic concepts
• reason about trade-offs
• read and understand code
• write code without constant AI assistance
• debug without copy-pasting prompts

Then something has gone seriously wrong in the process.

My advice is to use AI as a multiplier, not as a replacement. Let AI handle the tedious parts so you can focus on architecture, edge cases, and understanding why the code works. Read the output critically. Question it. Rewrite it when needed. Build things from scratch now and then to keep your skills sharp.

Technology is moving faster than ever, but human competence doesn’t seem to be keeping up. If we don’t start valuing real thinking again, instead of just prompts, shortcuts, or trends, we’ll end up building a future that looks impressive but rests on a weak and fragile foundation.

The engineers who will succeed aren’t just the ones who can prompt quickly, but those who can think for themselves when AI makes mistakes.