Jensen Huang is at it again. In his ongoing campaign to convince the world that human cognition is simply legacy software, the Nvidia CEO has reportedly embraced a new talking point: Software engineering is a "solved problem," and therefore, it is no longer a "smart" job. The really smart people, he implies, are the ones "discovering new problems" and letting AI agents handle the trivial, blue-collar task of actually building the solutions.
When a man whose net worth is inextricably linked to the mass consumption of AI hardware tells you that the craft of writing software is trivial, you have to sit back and admire the sheer audacity of the grift. It is the digital equivalent of a salesman selling you a $40,000 oven and confidently assuring you that "cooking is a solved problem—you just tell the oven what you want to eat."
The Illusion of the "Solved Problem"
If software engineering is a solved problem, someone forgot to tell the software.
To declare this discipline "solved" is to profoundly misunderstand what software engineers actually do. It assumes that coding is merely the act of translating a perfectly unambiguous set of requirements into syntax. But anyone who has ever built a production system knows that perfect requirements are a myth.
Engineering is not typing. Engineering is the agonizing process of discovering edge cases, negotiating systemic trade-offs, debugging race conditions in distributed systems, and asking the business logic questions that the product manager didn't even know existed. AI can autocompete boilerplate, and it can summarize documentation, but it cannot intrinsically understand why a particular microservice architecture is going to buckle under the specific, weirdly shaped load of your Black Friday traffic.
As we've seen with tools like SWE-CI, autonomous agents currently boast a 30% regression rate when asked to maintain a real codebase over time. Calling this a "solved problem" is like calling aviation a solved problem because you built a plane that only crashes three times out of ten.
The Devaluation of Implementation
The most insidious part of this narrative is the reclassification of "smart." By Huang's metric, the deep, structural understanding of how a computer actually works is no longer intellectually valuable. The new genius is simply the "Idea Guy"—the visionary who gestures broadly at a whiteboard and expects an LLM to magically spawn a scalable infrastructure.
We are witnessing the systematic devaluation of implementation. Tech leadership is desperately trying to convince itself that the messy, fraught, and highly skilled labor of maintaining complex digital infrastructure can be hand-waved away by buying more GPUs.
But when everyone is an "Idea Guy" and the people who know how the engine actually works are dismissed as obsolete, you don't get a utopia of rapid prototyping. You get a terrifyingly fragile ecosystem built on layers of hallucinated abstractions that nobody knows how to fix.
The Ultimate Convenience
Let’s be clear about why this narrative exists. It isn't a reflection of technical reality; it is a sales pitch. If software engineering is a manual, outdated trade, then corporate America's massive expenditure on AI compute is completely justified. Nvidia's valuation demands that you are obsolete.
You don't "solve" software engineering any more than you solve physics. You just shift the complexity. And when this massively complex, AI-generated house of cards inevitably starts to topple, we are going to find out very quickly just how "smart" it is to fire the only people who know how to rebuild it.