How will engineers survive in the age of AI?

My Honest Thoughts

I've noticed something odd about LinkedIn job postings lately:

Companies are asking for much higher specs than they used to, and yet very few entry-level positions are open. Most of the jobs are for contracting departments with university-industry connections, and the few openings that remain attract top-tier candidates. Career positions are not much different. There aren't many new positions per se.

While the total number of people hiring has decreased, the variety of roles and locations has increased. In the news, they say that they are hiring a lot of semiconductor workers, but in the field, even if you apply with a top-tier thesis, it is difficult to pass the paperwork. The temperature difference between news and reality is quite large.

Who is actually hired? In the end, it is the person with more practical experience in the field. It's not just the newcomers, it's the experienced ones.

Companies are asking us to use AI to increase productivity.

• But when you use it, it feels weird.
• Your output per unit of time has definitely increased with AI.
• But you don't feel that the total amount has exploded.
  • "I feel like I've done more than before..." This awkward feeling. Even if you ask your seniors, no one can answer you properly."

This is because it is a problem that our generation is experiencing for the first time.

In the interview room, there is still a hand-coding test.(Not pseudo code, real leet coding test) But the engineers in the field are actually working by organizing specification documents, typing them into prompts, and reviewing what the AI implements.

The interviewers are just doing what they've done in the past.

So, how do STEM people make a living in the age of AI?

"Doers" vs. "Takers"

I think this distinction is the most important right now.

People who suggest and decide what to build vs. people who implement it

AI is rapidly replacing people who implement things. Scripts that used to take hours to write now take minutes. Now it's people tell it to do it, AI drafts it, people review it."

AI is really good at implementation. It's good at recognizing problems, but it's not as good at implementing."

So what we're seeing in the field now is this structure:

• Product Owner synthesizes the customer's problem, defines it in a specification document, and hands it off to R&D engineers.
• The customer-PO-R&D triad gets together to discuss "what to design."
• What architecture will work for this application, what tradeoffs will be made, this is still the domain of humans.

As AI-driven development environments eventually take hold, the roles of the survivors have begun to divide up like this.

1. Problem recognition specialist - a keen eye for recognizing hard-to-recognize problems
2. Design judgment specialist - a big-picture thinker who understands technical tradeoffs and makes big architectural decisions
3. AI orchestrator - a managerial person who puts AI to work and monitors the results.
4. Director - A detail-oriented person who takes final responsibility for the product before it goes to market.

The first to go is the first to be replaced

The best and brightest around us. People with no taste, high GPAs, no creativity, good at memorizing, and good at being told what to do. These people will be the first to be replaced by AI, because that's exactly what AI is best at. Problems that have a set answer, problems that follow rules, problems that have a lot of data.

If you look at the models that AI has started to recognize problems, they've gotten better at recognizing problems as they keep asking themselves questions, as they've come up with models that are reflective, like deep think, thinking models.

The engineer who asks, "Is this specification the best? Is there something better? Why???" is more valuable than the engineer who designs according to the specification sheet provided.

One of the things I learned in my management classes was the ability to ask questions. If you know the technology, you can answer the "how" questions. But the "what" and "why" are often missed by engineers.

Solo entrepreneurship has become a realistic option

Twenty years ago, it was not easy to start a business on your own

• You needed a team
• You needed capital
• You needed to be an expert in your field.

A decade ago, the era of solo entrepreneurship slowly began:

• Development frameworks
• Development platforms

With these things, solo developers and solo game development were born.

Now it's one dimension bigger:

• Claude Code
• Codex
• Gemini
• Cursor

Open it up, assign roles, give it a deep research topic, give it an idea prompt, and in a few hours you have a prototype.

The development iteration has become insanely fast:

• If it's broken, I flip it,
• see user reaction, flip it again,

It used to be that I'd have a meeting with developers, designers, and data analysts, and I could see the direction. Now, I just make the decision and let the AI do it.

A 100-hour work week by myself, for $20/month and a few AI subscriptions, I can be as productive as a team.

Of course, there's still the hard part.

Discovering a customer problem, validating that it's a real problem, and positioning it in a way that fits the market - that's still human work.

And frankly, the combination of STEM skills and business is pretty powerful here. You know what's technically possible, and you know how to sell it business-wise.

I guess what I'm saying is, "consistency" is also a weapon

Whatever you build these days gets copied quickly. It's just a click away. So it's becoming more important to look at who is doing it:

• their story
• schools
• career
• papers, patents, awards
• presentations, side projects, ...

I'm learning this too.

My advice

1. Keep learning the technology, but don't be a slave to the AI tools, and have a solid foundation in recognizing when AI is wrong. Only then can you become a "teller."
2. Gain experience building things on your own. Beyond your day job, define a problem, create a solution, and put it out into the world. This is the muscle that will survive the AI era.
3. Find what you love: What you really like, what you care about. That's the hardest thing for AI to replace.
4. Document and communicate consistently No one will recognize your skills first. You have to make it known. Whether it's a paper, a blog, LinkedIn, or GitHub.

I don't have a finished answer. I'm just being honest because I don't want you to have to go through this alone.

If you're going through the same thing, let's talk.