Are We Trusting LLMs Too Much?

Lately, I have been thinking about the turmoil caused by LLMs and how seriously they have taken over the world — and the minds of people.

Many of us believe that LLMs are the future, and in some sense, they are right. Professional developers are no longer writing code entirely by hand. They give a spec, and tools like Claude or Codex produce clean, working code for most cases. It is impressive, no doubt.

But I am not trying to be pessimistic here. I am just trying to look a few years ahead and predict what our future might look like — at least from the perspective of a developer’s career.

The Hype Will Teach Us Hard Lessons

I believe the current hype around LLMs will eventually teach us some very important lessons. We are already starting to see the cracks:

• Security issues. When code is generated at scale without deep human review, vulnerabilities slip through. LLMs do not truly understand security context — they pattern-match from training data.
• Boring, cookie-cutter products. When everyone uses the same tools to generate the same kind of output, differentiation disappears. We risk a wave of products that all look and feel the same.
• Blind trust in output. Some of us are placing absolute faith in what LLMs generate — and that worries me.

When AI Becomes the Advisor for Everything

Recently, I unwillingly witnessed a situation that stuck with me. A young girl was discussing her personal relationship problems with an LLM, taking its recommendations seriously. The advice it gave was awful.

This is not about one bad response. It is about the pattern: people are increasingly turning to LLMs not just for code or search queries, but for life decisions. And LLMs are not built for that. They have no empathy, no context about your life, and no accountability for the consequences.

AGI Is Coming — But Not Tomorrow

I do believe that general AI will appear eventually. But it seems like it is not happening as soon as the hype suggests.

Here is how I see the trajectory of AI capability over time:

We had a sharp jump when LLMs broke through. Right now, we are somewhere around the plateau — capability is real, but the growth has slowed. I expect a period of stagnation and correction before we see the next real leap. Think of it as the classic hype cycle: breakthrough, inflated expectations, disillusionment, and then genuine, sustained progress.

So What Should We Do?

I am not saying stop using LLMs. I use them myself every day. But I think we need to treat them as tools, not oracles. Review the code they generate. Question the advice they give. And most importantly — keep sharpening the skills that LLMs cannot replace: critical thinking, judgment, and the ability to ask the right questions.

The developers who will thrive in the next decade are not the ones who blindly delegate everything to AI. They are the ones who know when to trust it — and when not to.

What do you think? Are we trusting LLMs too much, or am I overthinking it?