"Software is terrible" saved humanity? Entrepreneur refutes AI bubble theory: the Singularity won't arrive overnight

In response to Citrini Research’s widely circulated AI disaster predictions, serial entrepreneur John Loeber offers a completely different perspective: the entrenched inertia of bureaucratic systems, the low quality of existing software, and the enormous potential of American re-industrialization will ensure that the AI revolution does not overnight overthrow human society. This article is based on John Loeber’s piece “Contra Citrini7,” translated and written by Dongqu.
(Background: AI boom turns into economic catastrophe? Citrini Research warns of a “Global Intelligence Crisis” in 2028)
(Additional context: AI panic unemployment! Microsoft executive warns: Most white-collar workers will be automated within “the next 12-18 months”)

Table of Contents

Toggle

• The software industry has an “unlimited demand” for labor
• The turnaround opportunity of “re-industrialization”
• Moving toward abundance

In 2007, people believed that “peak oil” would mark the end of U.S. geopolitical dominance; in 2008, the dollar system seemed on the brink of collapse; in 2014, many thought AMD and NVIDIA were about to become history. Then ChatGPT emerged, and everyone declared Google finished… Yet history has shown that well-established institutions demonstrate far more resilience than outsiders expect.

When Citrini talks about institutional change and fears of rapid labor displacement, he writes: “Even those fields we think are maintained by personal relationships are vulnerable. For example, in real estate, buyers have tolerated 5%-6% commissions for decades because of information asymmetry between agents and consumers…”

Loeber couldn’t help but laugh here. “Real estate agents are about to disappear”—this has been shouted for 20 years! No need for super intelligence; Zillow, Redfin, or Opendoor are more than enough. But this example actually proves the opposite of Citrini’s point: despite being considered outdated by most, real estate agents are thriving more than anyone predicted ten years ago, thanks to market inertia and regulatory capture.

A few months ago, Loeber bought a property himself. The transaction required hiring an agent, supposedly for good reasons. His buyer’s agent earned about $50,000 from this deal, but the actual work—filling out forms and coordinating with multiple parties—took maybe 10 hours, which he could have done himself. The market will eventually become more efficient, with labor prices adjusted accordingly, but that will be a very slow process.

Loeber has firsthand experience with inertia and change management: he founded and sold a company that helped insurance brokers shift from “manual service” to “software-driven.” He learned that—despite the enormous progress—human social systems in the real world are far more complex than imagined, and everything takes longer than expected, even when accounting for this law. This doesn’t mean the world won’t undergo drastic change; it means change will be gradual, giving us time to adapt and respond.

The software industry has an “unlimited demand” for labor

Recently, the software sector has underperformed because investors worry that backend systems of companies like Monday, Salesforce, and Asana lack moat and are easily copied. Citrini and others believe AI programming signals the end for SaaS companies: products become homogeneous and unprofitable, and job opportunities evaporate.

But everyone overlooks one thing: these software products are actually terrible.

Loeber claims this from experience, having spent hundreds of thousands of dollars on Salesforce and Monday. Yes, AI can enable competitors to copy these products, but more importantly, AI allows them to make better ones. Stock prices falling isn’t surprising: an industry long reliant on bundling sales, lacking competitiveness, and filled with low-quality legacy vendors is finally facing real competition.

From a macro perspective, almost all existing software is garbage—this is a well-known fact. Every paid tool is riddled with bugs; some are so bad that you can’t even pay for them (Loeber has been unable to use Citibank’s online banking for transfers over the past three years); most web apps fail to properly support mobile and desktop; no product fully meets your needs. Silicon Valley darlings like Stripe and Linear are popular mainly because they’re not as frustratingly difficult to use as competitors. Ask a senior engineer for a “truly perfect software,” and you’ll likely get long silence and awkward looks.

There’s a deeper insight here: even if we reach a “software singularity,” human demand for software labor is nearly infinite. It’s well known that the last few percentage points of perfection require the most effort. By this standard, almost every software product has at least 100 times more complexity and features to develop before reaching saturation.

Loeber believes that most critics claiming the software industry is about to die lack practical software development intuition. The industry has existed for 50 years; despite huge advances, it’s always in a state of “supply shortfall.” As a programmer in 2020, his productivity was equivalent to hundreds of programmers in 1970—an incredible leverage—but there’s still enormous room for optimization. People underestimate the power of the Jevons Paradox: efficiency improvements often lead to explosive growth in total demand.

This doesn’t mean software engineering is an eternal safe job, but the industry’s capacity and inertia to absorb labor are far greater than imagined. The saturation process will be very slow, giving us ample time to prepare.

The turnaround opportunity of “re-industrialization”

Labor shifts will inevitably occur, such as in driving. As Citrini notes, many white-collar jobs will indeed experience upheaval. For roles like real estate agents, who have long lost substantive value and rely entirely on inertia, AI might be the final blow.

But the opportunity for revival lies in America’s near-infinite potential and demand for re-industrialization. You may have heard of “manufacturing returning,” but it’s much more than that. The U.S. has largely lost the ability to produce key components of modern life: batteries, electric motors, small semiconductors—the entire power industry chain depends heavily on overseas supply. What if a military conflict breaks out? More critically, China produces 90% of the world’s synthetic ammonia—if supply is cut, fertilizer can’t be made, leading to famine.

Focusing on the physical world reveals endless job opportunities. These are foundational projects that benefit the nation and create jobs, gaining bipartisan political support.

The economic and political winds are shifting in this direction—discussions of manufacturing resurgence, deep tech, and “American vitality.” Loeber predicts that as AI impacts white-collar sectors, the most logical political response will be large-scale re-industrialization, employing “massive employment projects” to absorb labor. Fortunately, the physical world doesn’t have a “singularity”; it’s constrained by real-world resistance.

We will rebuild bridges and roads. People will find more fulfillment in tangible labor than in digital abstractions. The Salesforce senior product manager who lost $180,000 annually might find a new stage at a “California desalination plant” to help end a 25-year drought. These facilities must be built to the highest standards and maintained long-term. If we’re willing, Jevons’ Paradox applies just as well in the physical realm.

Moving toward abundance

The end goal of large-scale industrial engineering is prosperity. The U.S. will regain self-sufficiency, achieving large-scale, low-cost production. Surpassing material scarcity is key: in the long run, if AI truly causes most white-collar jobs to vanish, we must be capable of maintaining a high-quality standard of living. Since AI will drive profit margins toward zero, consumer goods will become extremely cheap, and this goal will be naturally achieved.

Loeber believes different sectors of the economy will “take off” at different speeds, and nearly all transformations will be slower than Citrini predicts. He clarifies that he is highly optimistic about AI and foresees a day when his own labor will become obsolete. But this takes time, and that time gives us the opportunity to craft good policies.

In this regard, preventing the market collapse Citrini describes is actually quite manageable. The U.S. government’s response during the pandemic proved its ability to act decisively in crises. When needed, large-scale stimulus policies can be quickly implemented. While acknowledging that government efficiency is often lacking, that’s not the key point. The focus should be on maintaining material prosperity for the people—a universal welfare that sustains the legitimacy of the state and social contract, rather than clinging to outdated accounting metrics or economic doctrines.

If we stay alert and adaptable during this slow but certain technological transformation, we will ultimately navigate it safely.