The Two Modes Every AI-Era Engineer Needs

The Information recently published a piece about "cracked engineers," a new archetype for the AI era. The cracked engineer is young, tireless, technically deep, and uses AI coding tools to do the work of entire teams. Startup founders are desperate to hire them. One founder dismissed all his intern applicants because "none of them would be cracked enough." Another's friend quit learning to swim because "it ate up too much of my productivity time."

I spend most of my working hours with Claude Code open. I run a personal AI assistant, a full infrastructure stack, and multiple services off a single VPS. I've lived the compression the article describes: one person doing what used to take a team. And the article captures something real.

But it gets the narrative wrong.

The cracked engineer is real

AI coding tools have genuinely changed what's possible for a single engineer. I can spin up a new service, wire it into my reverse proxy, and have it live in an afternoon. Tasks that used to take days of boilerplate now take minutes. The floor has risen and the ceiling has risen with it.

The article is right that this shift rewards a certain kind of engineer: someone with deep technical skills who embraces AI tools rather than scoffing at them. Someone versatile enough to work across the stack. Someone willing to put in serious hours.

Where it goes wrong is turning this into an identity. "Cracked" isn't a personality type. It's a way of working. And the way the article describes it (all gas, no brakes, delete your hobbies) misses the rhythm that actually makes long hours productive.

Mode one: push

When you're deep in a hard problem, hours are everything. Not as a flex, but because complex systems demand sustained immersion. You need weeks of consecutive focus on the same problem to hold the full picture in your head. Context-switching kills this. You can't solve a hard architecture problem in 45-minute blocks between meetings.

This is where the "cracked" label is earned. Multiple AI sessions running, fast iteration, shipping aggressively. The article mentions an engineer who built a prototype for fine-tuning large AI models in a few weeks with help from AI tools, accomplishing what some open-source efforts took a full year to do. That tracks. I've had similar stretches where everything clicks and you're producing at a rate that would've seemed impossible a few years ago.

Push mode is real, it's powerful, and it requires long hours. No argument there.

Mode two: evaluate

Here's what the article doesn't talk about.

After days or weeks of pushing, you need to step back. Not because you're tired (though you might be), but because you need a different kind of thinking. Push mode is about execution. Evaluate mode is about judgment.

Is this the right architecture? Is the AI-generated code carrying hidden debt I haven't noticed because I've been moving too fast? What does this system look like in six months when I need to maintain it? Am I even solving the right problem?

These questions can't be answered mid-sprint. They require distance. They require the kind of thinking that happens when you go for a walk, sleep on it, or, yes, learn to swim.

The guy in the article who quit swimming didn't gain productive hours. He lost the space where evaluate mode happens. Your brain doesn't stop working when you step away. It reorganizes. Every experienced engineer knows the feeling of waking up with the solution to yesterday's problem. That's not magic. That's your brain doing evaluate mode in the background.

The rhythm between them

The actual skill isn't living in one mode. It's knowing which mode you're in and when to switch.

Push when the problem is yielding. When you're in flow and the pieces are falling into place. When you can feel the momentum. These are the stretches where you work 12-hour days and it doesn't feel like grinding because you're making real progress.

Evaluate when you notice yourself brute-forcing. When the architecture feels wobbly. When you've been heads-down so long that you've lost sight of why you're building what you're building. When you catch yourself generating more code without stopping to read what you've already generated.

AI tools make push mode incredibly powerful. But they also make evaluate mode more important than ever. You're generating more code, making more decisions, creating more surface area for subtle mistakes. The speed is a gift, but it means the cost of building the wrong thing, or building the right thing the wrong way, is higher than before.

What AI tools actually demand

The article says cracked engineers "catch errors in AI-generated code." True, but undersold. The real demand is judgment:

Which of the five approaches the AI suggests actually fits your system? When does something look correct but will break in production? How does a change ripple through your stack? When is the AI confidently wrong?

These skills don't come from grinding harder. They come from experience and reflection. From building things, maintaining them, watching them break, and understanding why. From having the space to think about what you're doing, not just doing it faster.

Strategy is the bottleneck now

The best quote in the article comes from a recruiter: "A cracked engineer is a Band-Aid for a lot of founders who have no idea what they're building."

AI has made execution so fast that knowing what to build is now the scarce resource. An engineer in permanent push mode building the wrong thing just arrives at failure faster. Evaluate mode is where you course-correct before wasting a week of cracked-level output on something that doesn't matter.

The hardest part of any project I've worked on was never writing the code. It was deciding what to build and why. AI tools haven't changed that. If anything, they've made it more acute.

The rhythm is the skill

The two modes aren't a compromise. They're what makes the long hours actually count.

Push hard when the problem is yielding. Step back when you need clarity. The engineers who will thrive in the AI era aren't the ones who deleted swimming from their lives. They're the ones who know that the walk home from the pool is where the next breakthrough starts.

*Malte Meng is CTO & Co-Founder at Bunka.ai.*