Post-AI: The Human Question, Revisited

---

The Broken Suit

I was rewatching Iron Man 3 recently when I stumbled upon a scene that perfectly mirrored the questions I'd been wrestling with.

Tony Stark, stranded in an unfamiliar town. His cutting-edge suit shattered to pieces, JARVIS offline. He spirals into a panic attack — because without the suit, he feels powerless. Then a kid says to him: "You're a mechanic. Build something."

That one line snaps him back. He goes to the local hardware store and rigs together weapons from off-the-shelf parts to take down the villain's stronghold. Not with cutting-edge tech, but with the ability to understand a problem and craft a solution from whatever's at hand. That was the essence of Tony Stark.

Watching that scene, the same question hit me. If you stripped away every AI tool I work with daily, what would be left?

Without the Suit

In The Avengers, Captain America asks Tony: "Take away the suit — what are you?" Tony answers: "Genius, billionaire, playboy, philanthropist." An inventory of identity that exists independent of the armor.

I'm an AI engineer. Every day I open Claude Code, accept Copilot's suggestions, call the GPT API. These tools are my suit. They write code fast, spot patterns, handle repetitive work. But what if all of it vanished? What exactly am I capable of on my own?

I stalled on this question for a while. Then a failure from a few years back resurfaced.

The "Why" I Never Asked

A factory came to us with a request: build an AI solution to automatically detect surface defects on their production line. We collected data, trained a model, delivered results.

The problem came next. When floor workers evaluated the system's detections, each person had a different standard for "defect vs. acceptable." What one inspector flagged as defective, another passed as fine. A model trained on inconsistently annotated data was never going to work. The project failed.

Looking back, the factory's real problem was that "the definition of a defect was ambiguous." If I had started by asking "Why do you want this solution?" we would have defined defect criteria before ever collecting data. The question should have come before the technology.

The lesson was clear: no matter how powerful the AI tools, they're useless if you don't ask "why." And no tool can ask "why" on your behalf.

What Remains Beyond the Tools

Tony Stark could build weapons from hardware-store parts because he had a fundamental understanding of physics. Tools change; principles don't. Likewise, there are things that remain after you strip the tools from an AI engineer.

The ability to define a problem precisely. The ability to read the gap between what's technically possible and what's actually needed. The sense that identical data and algorithms can yield completely different value depending on context. These aren't autocomplete features — they're earned through countless failures and iterations.

This isn't just an AI engineer's story. Take Figma away from a designer and they still have the instinct for "what users actually want." Remove a marketer's automation tools and they still know "why this message moves people." Strip an accountant's AI analytics and they can still judge "what this number means for the business."

Tools accelerate the "how." But it's people — not tools — who decide the "why" and "what." The faster AI takes over the "how," the sharper the human role of asking "why" and reading context becomes.

The Next Question

"Will AI replace my job?" Many people are anxious in front of this question. But I think the question itself is aimed in the wrong direction. Asking whether you'll be replaced is like worrying whether the suit will break. The suit always breaks eventually. What matters is what comes after.

Tony Stark was a mechanic even without the suit. His essence wasn't the armor's specs — it was being the person who solves problems with whatever's in front of him. The same goes for us. Tools will keep changing. Today's cutting edge is tomorrow's legacy. But the ability to ask "why should we do this?", to judge within context, to set direction under uncertainty — that doesn't belong to the tool. That belongs to you.

So let's reframe the question:

When AI takes over the work you've been doing, what value will you create with the time it frees up?

If you can answer that, you'll be fine no matter which suit breaks.