Navigating the AI Revolution: Who Wins, Who Loses, and How to Prepare | Emad Mostaque

Emad Mostaque argues that AI has reached a takeoff point: systems are becoming more capable and consistent, and he expects human cognition’s economic value to turn negative in about two years. He points to agents that can book restaurants and check school activities, as well as advances in mathematics, 3D scenes, music, web design and video. He says OpenAI used 10,000 agents for 88 hours on a difficult mathematical problem, an effort he compares with 100 years of talented human work. A reported Navier–Stokes proof, however, meets only two of the problem’s four qualifying conditions, remains under review and is not a solution to the full Millennium Prize Problem.

Mostaque sees procedural, computer-mediated work as especially exposed, with disruption likely to arrive abruptly rather than steadily. He cites a call-center owner expecting competitive pressure to force a 70,000-person workforce down to zero over the next few years, while emphasizing that the owner does not want to fire workers. He predicts white-collar job losses will exceed 10% by 2029 and accelerate afterward, and says well-trained AI already outperforms entry-level workers in many digital jobs. The host describes a game project shrinking from more than 100 people working on it at once to six or seven, but cautions that the original team was inexperienced and not all contributors worked full-time.

Cost, speed and workflow integration may matter as much as further gains in intelligence. Mostaque says a model-adaptive harness helped the cheapest DeepSeek model outperform OpenAI’s then-frontier model at 50 times lower cost, and predicts equivalent performance could become 100 times cheaper within a year as chips and harnesses improve. He says AI still takes at least a year or two to integrate into workflows, and cites an Anthropic Economics Department scenario projecting 15% annual GDP growth by 2030 alongside a 12% reduction in the cognitive-labor workforce needed. AI swarms can outperform individual use, but he notes a reported $500,000 coding bill as evidence that high-end models remain costly.

Mostaque expects robots to take on more human-shaped work, but says production limits will constrain how quickly they spread. He estimates a capable humanoid could perform about five times a human’s labor at a cost of roughly $30,000 per year; he considers one billion robots by 2030 optimistic, with hundreds of millions over a ten-year horizon more plausible. He also anticipates a major infrastructure buildout, including data centers and robotics, even as labor displacement makes its benefits uncertain. His proposed response is to learn AI through daily use, test which parts of a job involve automatable information-classification tasks, and become more capable and flexible.

As selling labor becomes harder, Mostaque argues, ownership may be a more durable route to income and wealth. He favors broad ownership of robots and data centers over relying on universal basic capital tied only to AI companies, and proposes locally owned AI companies funded by institutions, pension funds, wealthy individuals and, where permitted, retail investors. He suggests reserving 10% of initial equity for children under 18 and allocating half a percent each year to every child born in a state. He rejects tax-funded UBI as unworkable at current scale, estimating US tax receipts at $5 trillion versus $5.1 trillion to provide $16,000 annually per person, while allowing that other mechanisms could be considered.

Mostaque warns that AI could concentrate power and weaken people’s control over decisions, especially as personal assistants gain access to accounts, schedules and other sensitive information. He argues that “sovereign AI” requires local ownership of the data centers and systems, not just locally branded services, and calls for consequential systems such as judicial AI to disclose their norms and data. He proposes an “AI union” that pools ownership, supplies citizens with agents, trains deployment engineers and takes stakes in data centers and robots. He also says people should own a stake in a personal AI interface and be able to remove it from the system.

The speakers disagree over whether cheaper intelligence, energy and labor will make goods broadly abundant or leave capital owners controlling access to scarce resources; they agree that some things, such as a home in a particular location, will remain scarce. Mostaque says society should guarantee a baseline standard of living while preserving the ability to advance through effort, and warns that an owner class and a deprived majority would be unstable. He expects job displacement to threaten purpose and status as well as income, and suggests a citizen service corps could provide young people with service and a sense of progress. His central message is that communities, ownership and preparation matter now: he expects the “sand pile” to begin collapsing in about two years.