Last year, news outlets ran story after story about how ChatGPT was killing entry-level jobs. Junior workers were done for. Then hiring data showed the entry-level market bouncing back, and the narrative flipped. Now AI supposedly creates jobs by helping companies grow. As Georgetown professor Cal Newport puts it, AI is "simultaneously contracting the job market for recent college graduates while also expanding the job market for recent college graduates."
He extends this critique to the whole AI industry.
In a recent essay drawing on Elizabeth Lopatto's reporting in The Verge, Newport argues that Silicon Valley abandoned customer problem-solving for "inventing the future." LLMs fit this pattern. So did NFTs. The metaverse, too. All solutions hunting for problems.
Of these, large language models have real potential. But Newport isn't letting AI companies off the hook. Most people use ChatGPT as a slightly more talkative Google, or to format an itinerary. Useful? Sure. Life-changing? Not yet. Meanwhile they get bombarded with breathless hype about GPT 5.5 benchmarks and terrifying predictions about AI destroying everything. Nobody cares that GPT 5.5 underperformed Opus 4.7 on SWE-Bench Pro. They want to know when someone will make a product that actually helps them.
The enterprise side is the same story. Goldman Sachs published research in 2024 questioning whether generative AI infrastructure costs will ever deliver returns comparable to past tech revolutions. Companies keep pouring billions into AI tools that mostly automate low-level admin work. The high-value gains remain theoretical. Newport's point is dead simple: until AI companies can point to something concrete that makes a normal person's day better, stop shouting about the future. Build something people asked for.