Peter Diamandis: Hopes Millions Will Use AI to Solve Specific Problems
Peter Diamandis stated that if artificial intelligence makes starting companies easier, he hopes to see millions tackling very specific issues: better tools for nurses, tools for elderly parents, and software that no one has thought of yet.
In the same week, he wrote that the cost of queries reaching GPT-3.5 level performance has dropped from about $20 per million tokens in November 2022 to about $0.07 in October 2024, a reduction of about 280 times; for the price of a cup of coffee, one can drive a prototype that five years ago required a team of experts or millions of dollars. A Pew survey shows that 65% of American workers rarely or never use artificial intelligence at work. He also cited Salim Ismail's question to CEOs: Is there a high-profit business in your company that can be replicated by two people with an agent in 60 to 90 days? 70% answered yes.
He has long compared the cost of entrepreneurship, writing that it was about $40,000 in 2005 and is expected to be about $110 in 2026. XPRIZE drives civilian teams to achieve goals considered impossible; Fountain Life focuses on health extension testing and intervention. He estimates that globally there are about 3.3 billion households, and by the end of 2027, about 700 million people may need home care, with insufficient supply of nurses, and discusses home care supplemented by humanoid robots.
"Organizational Singularity" is a term he and Ismail use together: the number of employees in companies could shrink to about 20% of current levels, and the number of startups could multiply; one person with an agent could orchestrate traditional staffing. He also separates longevity escape velocity, care companionship, and "cure does not equal companionship": models can diagnose, but being present is still a human job, and tools should serve nurses and family members rather than replace presence.
In market mechanisms, the buyers are nurses, caregivers, and small entrepreneurs who hold specific scenarios, while the sellers are exponential technology evangelists who turn cost-reduction narratives into calls to action. Funding flows from speeches, memberships, and longevity healthcare into vertical small software, rather than a single app store hit. Events are driven by postings. Beneficiaries are creators who can turn shifts, medication, and family notifications into small tools; those under pressure are high-profit departments or intermediate software that still rely on staffing and distributors, which can be replicated by two people within 90 days.
He chooses to stand on the side of "building with AI" rather than just discussing job losses.
Source: Public Information
ABAB AI Insight
Diamandis summarizes the previous "complaints equal demand" list: nurses, aging parents, and uninitiated software. This is not a new discovery of the care gap, but rather a reduction of XPRIZE-style incentive logic to the individual level: grand prizes for space, small tools for shifts and care. Fountain Life sells affordable longevity packages, while postings sell unproductized family scenarios, both pointing to the willingness to pay due to an aging population.
Capital spills over from the price reduction of general large models. After token costs drop two orders of magnitude, the marginal development cost of vertical software approaches that of prompts and acceptance, no longer needing to pass through venture capital committees first. 70% of CEOs admit that internal profit centers can be replicated by two people, effectively giving external teams a map for attack. The nursing workforce gap provides real budget sources: families, insurance, and healthcare institutions, rather than creating another general assistant.
The benchmarks are early electronic medical record giants that turned hospital processes into suites, and numerous health applications that failed to penetrate nursing stations. The industry is in an expansion phase where general models are sufficient, but vertical workflows have not yet been filled. Whoever first turns "medication checks, shift changes, and family schedules" into a subscription tool will price by hours, rather than by model parameters.
Structural changes belong to technological substitution. The long cycles of software in hospitals and nursing homes are interrupted by weekly prototypes from individual developers, provided that responsibilities and healthcare interfaces remain slow. The mechanism is: models make interfaces cheap, while regulation and responsibility slow penetration, so the first batch of money will fall on small tools that nurses themselves are willing to install, rather than replacing core medical record systems.
ABAB News · Law of Cognition
- After models become cheap, what is valuable is the specific pain that no one has initiated.
- Profit centers protected by staffing are most afraid of two people plus one agent.
- The nursing gap will not wait for software companies to finish product development before appearing.