Stop Coding, Start Supervising
Last updated on
Stop Coding, Start Supervising

Back to journey
January 15, 2026 4 min read journey
#AI #Software Engineering #Nvidia #Jensen Huang #Prompt Engineering #Responsible AI #Developer Identity #Creative Destruction #Future of Work #Old Dog New Flex #AI Agents #Tech Trends 2026

Stop Coding, Start Supervising: Nvidia’s CEO Just Redefined What It Means to Be an Engineer

On January 15, 2026, Nvidia CEO Jensen Huang told a crowd of engineers and developers at CES what may become one of the most quoted phrases of the AI era:

“I don’t want you to write code anymore. I want you to supervise the code that’s written.”

Let that sink in.

Coming from the head of the world’s most valuable chipmaker, whose products power nearly every major AI model on the planet, this wasn’t just a throwaway soundbite—it was a rallying cry.

And it signals a dramatic, perhaps uncomfortable truth:

The job of software engineering is being redefined.


From Coders to Conductors

What Huang is pointing to is something I’ve felt for a while—and something many engineers I’ve spoken to are grappling with, too: the role of the developer is shifting from craftsperson to orchestrator.

AI tools like GitHub Copilot, Claude Cowork, and Cursor aren’t just assisting—they’re producing production-grade code. They’re building files, importing dependencies, solving bugs, and even writing tests. Developers are increasingly stepping into a supervisory role:

  • Prompting models
  • Validating outcomes
  • Reviewing generated code
  • Coordinating agent workflows

The skillset is changing. Deep expertise in syntax is becoming less critical than:

  • Model intuition
  • Prompt design
  • Behavioral debugging
  • Outcome validation

Why This Moment Matters

Jensen Huang’s statement breaks a long-standing taboo: that coding—the sacred act of our discipline—may no longer be the center of gravity.

And it’s not just Nvidia. Companies are:

  • Hiring engineers for their ability to think and judge, not just type
  • Testing aptitude and reasoning over language-specific skill
  • Building agentic workflows where codebases evolve through AI assistance

This isn’t theoretical. It’s happening now:

  • A team at Cursor built a working browser from scratch using GPT-5.2 agents
  • HackerEarth reports a 50% rise in logic-based hiring tests over coding ones
  • Developers admit they don’t fully trust AI code—but they still ship it

The Engineering Identity Crisis

For many of us—including me—coding is more than just a job. It’s part of our identity. We learned it the hard way. We debugged by fire. We took pride in the grind.

So being told not to code? It feels like someone ripped out the engine behind our title.

But what if this isn’t the end of engineering?
What if it’s the beginning of a more strategic, creative, and impactful role?

When code is handled by AI, what’s left is:

  • Asking the right questions
  • Defining the right outcomes
  • Validating ideas
  • Teaching machines how to think like us

What You Can Do (Besides Panic)

This shift doesn’t mean you’re obsolete. It means you’re being up-leveled.

Here’s how to lean in:

  1. Practice prompt engineering — treat it like a DSL for managing behavior
  2. Review AI-generated code often — build intuition for model quirks
  3. Think in terms of outcomes, not diffs — your job is to define success
  4. Start documenting for humans and models — write for your future AI teammates
  5. Mentor models like junior devs — train, test, and verify them

Final Thought: This Isn’t the End—It’s the Promotion

Engineers aren’t being replaced. We’re being asked to level up.

To become supervisors, curators, strategists.
To teach machines, not outcompete them.
To lead the orchestra instead of playing every instrument.

And this isn’t just happening in software.

My friend Jason Sonderman wrote a fantastic piece back in June 2025 about AI and the creative industry titled “Design is Dead: Or How I Learned to Stop Worrying and Love Creative Destruction”.

In it, he captures a familiar emotional arc—first resistance, then uncertainty, then reinvention. The roles of designers, like developers, are changing fast. The value isn’t in drawing pixels or writing code—it’s in the decisions, judgment, and taste we bring to the table.

So yeah—maybe you’ll write less code in 2026.

But the things you build?

They might finally be bigger than what you could code on your own.


Want more AI-software commentary like this? Follow along at olddognewflex.com

Old Dog, New Flex Logo