So, for all the wonders of our “AutoEVERYTHING” world, science and technology will always need the outsiders, the artisans, the dreamers who challenge the status quo for the sheer love of the craft. They’re the ones who will move the world, not the AI.

Here’s the core difference: AI will never feel pride, obsession, curiosity, or that endless itch to create something brand new just because: “Why not?”
Progress depends on that kind of madness.

There will always be people who can’t settle for the “official pipeline,” who would rather battle limitations than settle for the possible, who write code as if sculpting wood, just to see if the result nudges the world even a little bit forward.

Let’s be honest, they’ll probably do the job of a fair share of software engineers before you know it). But there’s one territory AI will never conquer: that irrepressible urge to push beyond the known.

Companies are fighting tooth and nail for the rare talent who understands not just how to use the tools, but how and why they work (or break). And yes, LLMs will keep improving and automating things that seem like magic today.

Despite all this extreme automation, the elusive handful of humans capable of designing models from scratch, making critical architectural decisions, optimizing performance down to the last FLOP, and actually moving the field forward are more valuable now than ever.

And then came LLMs, pushing absurdity to new heights: a single model classifies, summarizes, labels, finds correlations—even makes your PowerPoint slides—all without breaking a sweat. Most of the time, you don’t even need frameworks anymore.

It was the era of hand-crafted architectures, homegrown models, and trial-and-error without a safety net. Today, most of that seems like ancient history. Frameworks, pre-built pipelines, AutoML: with just a couple of clicks, anyone can spin up a classifier or model tabular data.