So there’s something interesting happening in the recent AI singularity debate. Many people now believe the transformation will happen within months or years, but I see something different: system inertia that runs far deeper than anyone imagined.

Look, I’ve seen this happen again and again. In 2007, everyone said America was finished because of peak oil. In 2008, people said the dollar would collapse. In 2014, people were convinced the GPU makers were done. Then ChatGPT arrived, and everyone panicked that Google would go extinct. But every time, big institutions with massive inertia turned out to be far more resilient. They move slowly, but they don’t die.

Take real estate agents as an example. People have been joking around saying this profession will disappear in 20 years! But Zillow, Redfin, and technologies that should make them obsolete already exist. Yet what’s the reality? Real estate agents are still alive with 5–6% commissions because of regulatory capture and market inertia. I bought a house a few months ago and was forced to use an agent. That agent only worked 10 hours but earned 50 thousand dollars. I could do that job myself, but the system doesn’t allow it. This proves that singularity is a concept whose complexity in the real world is often overlooked.

Now, about software. People worry that AI will make programmers unnecessary. But they forget one thing: almost all software today is garbage. I’ve spent hundreds of thousands on Salesforce and Monday—both platforms are full of bugs and incomplete features. Web apps can’t adapt to mobile. My Citibank internet banking couldn’t transfer money for 3 years. Companies like Stripe and Linear only look good because their competitors are worse.

The key point is this: even if AI lets competitors copy products, AI also lets people create better products. The demand for improvement is almost unlimited. Current standards are still far from perfect—every piece of software still has room for 100x improvement before saturation. Programmers in 2020 can be as productive as hundreds of people in 1970, but the results still have a lot of gaps. This is the Jevons paradox: efficiency often leads to an explosion in demand, not a reduction.

Next is the issue of job displacement. Yes, it will happen. But the solution is reindustrialization. America has nearly completely lost its production capacity: batteries, motors, semiconductors, synthetic ammonia. China produces 90% of the world’s ammonia. If the supply is cut off, there’s no fertilizer—there’s famine. This is an unlimited job opportunity that builds real infrastructure and has bipartisan support.

My prediction: when AI disrupts white-collar workers, the government will fund large-scale reindustrialization as the path with the least political obstacles. We’ll repair bridges and roads, and build seawater desalination facilities. A Salesforce product manager earning 180 thousand could find a new job in infrastructure projects. The physical world has no singularity—it’s subject to real friction and real resistance.

Long term? Abundance. If truly many office jobs disappear because of AI, we have to be able to maintain a high quality of life. But since AI pushes profit margins toward zero, consumer goods will become super cheap. That goal will happen automatically. The transformation will be slower than doomsday predictions, and that gives us time for good strategies. The U.S. government has already proven it can respond quickly during crises—large-scale stimulus can be applied immediately if needed. The point isn’t accounting efficiency; it’s ensuring material prosperity and maintaining the social contract. If we stay alert and responsive, we’ll be safe. AI disruption will happen, but not in a single day.