Y Combinator has published its Fall 2026 Requests for Startups, with a central theme that cuts across nearly every category: AI is moving into the physical world.
The accelerator said generative AI is no longer limited to chatbots or coding tools. In its view, the technology is beginning to reshape education, healthcare, finance, defense, energy, infrastructure and the way humans work.
An AI tutor for every child
One of YC’s education ideas is called The Primer, inspired by the science fiction novel The Diamond Age.
YC described the fictional Primer as more than an electronic book. It is an AI mentor that stays with a child for years and adjusts its teaching style as the child grows. The firm argued that the best education in history has often taken the form of one-on-one tutoring, but that model has only been available to a small group of students. With large language models maturing, YC said it is now possible to imagine every child having an AI teacher that is patient and tailored to the way that child learns.
YC said it wants to see AI education products that help young children learn reading, writing and math, while continuously adapting materials to each student’s level. The goal, it wrote, is not to replace teachers but to bring something closer to private-tutor-quality education to more students.
U.S. Army Secretary makes a direct call for startups
The most unusual piece in this year’s RFS came from U.S. Army Secretary Daniel P. Driscoll. According to ABMedia, this is the first time a sitting U.S. cabinet-level official has directly contributed to the RFS and openly asked founders around the world for defense technology solutions.
Driscoll wrote that modern warfare is changing faster than the traditional military procurement process can handle. The Army, he said, needs far more commercial technology instead of waiting years for new weapons to be developed.
YC listed the areas where the U.S. Army is most interested in startup participation:
- Low-cost intercept systems
- Drones
- Next-generation sensors
- Modular military hardware
- Military logistics
- Equipment for extreme environments
- Advanced manufacturing
Driscoll also said that if a product shows promise, the Army is willing to provide funding and testing grounds to help deploy the technology more quickly. ABMedia said the move shows U.S. defense tech is embracing startup culture on a broad scale.
Cloud infrastructure built for small software
YC also highlighted software infrastructure through a category called A Cloud for Small Software.
Its argument is that AI agents have made it much easier for individuals to build small pieces of software, but deployment remains hard. Large cloud platforms such as Amazon Web Services and Microsoft Azure were designed for services with millions of users. For tools with only a handful of users, YC said, those systems are too complicated.
The firm wants someone to build what it called a “Small Software Cloud,” where sharing an AI tool feels as simple as sharing a Google Doc, while still handling permissions, security and software deployment.
From single-user chat to collaborative AI
In its Multiplayer AI category, YC argued that AI should not remain a one-person chat interface.
It pointed to products such as Google Docs replacing Word and Figma moving past Photoshop, saying the common thread was collaboration. YC expects future agents to behave more like coworkers, letting multiple people watch an AI system work, edit its output together, hand off tasks and finish work jointly.
The examples it gave include engineers coding together, sales teams working with customers together, lawyers revising contracts together and analysts building models together. Those tasks, in YC’s view, should be completed by teams working with AI rather than individuals using AI alone.
AI compute at sea
YC also floated the idea of Compute at Sea. Its position is that the main bottleneck for AI is no longer the GPU by itself, but land, power and cooling.
The firm noted that the ocean covers 70% of the planet, does not require large amounts of land, offers natural cooling, can pair with solar energy and may face less local government resistance. On that basis, YC suggested that large offshore fleets of AI compute ships could emerge as a new layer of global cloud infrastructure.
Consumer AI and the aging population
For consumer products, YC said the web era produced Google and the mobile era produced Instagram, yet the AI era has not produced a true consumer giant of the same order.
It argued that the issue is not demand but cost. AI used to be too expensive. Now, model quality is approaching human level and token costs are falling quickly. YC said that opens the door for a real wave of consumer AI products across health, entertainment, learning, transportation, finance and social interaction.
Another focus area is AI for the Aging Population. YC wrote that by 2030, one in five people in the United States will be older than 65, while care labor remains in short supply.
It outlined several areas where AI could help:
- AI companionship
- Voice assistants for older adults
- At-home monitoring
- Care robots
- Family care management platforms
YC described this as one of the largest long-term markets globally.
Rebuilding the operating system for physical work
YC also looked beyond office work. It wrote that 80% of the global workforce does not sit in an office, and industries such as construction, logistics, repair and manufacturing are still running on management systems designed two decades ago.
That no longer fits the labor mix YC expects in the future. It said workers will increasingly include humans, AI agents and robots, and it wants founders to build new operating systems that can manage all three. In that setup, companies would manage not just employees, but AI systems and robots as well.
YC says crypto downturns are still a time to build
Crypto also appears in this year’s RFS. In a section titled The Best Time to Build in Crypto, YC partner Nemil Dalal said now is the best time to start a crypto company.
Dalal argued that while market conditions are weak, the underlying infrastructure is maturing quickly. YC specifically highlighted the following areas:
- Capital raising
- New stablecoins and stablecoin applications
- Agentic commerce
- Trading
- Institutional products
- Scalable and private blockchains
YC also predicted that nearly all startups will eventually use crypto, often without realizing it, as part of their payments and financial infrastructure.
Real-world data, proof of personhood and compliance
Several other categories focused on infrastructure.
In Data for the Real World, YC said AI can already understand text and images, but still lacks large volumes of real-world data. It wants more companies to use robots, balloons, sensors and autonomous devices to collect data from industries such as energy, agriculture, construction and logistics. The firms that control those datasets, YC argued, may be able to build AI models for the physical world itself.
In Proving You’re Human, YC said AI deepfakes are driving a rapid rise in video and voice fraud. It cited a case in which a finance executive transferred as much as $25 million after attending a meeting populated entirely by AI-generated people. For that reason, YC argued that one of the most important infrastructure layers of the next decade will be a new system for verifying that someone is a real person. Banks, social platforms, dating apps and even all video meetings could end up needing it.
In AI-Native Compliance Infrastructure, YC said compliance remains one of the most expensive functions inside companies. It argued AI can track regulations in real time, build audit records, generate reports automatically and detect anomalies, effectively rebuilding compliance infrastructure for global businesses. Multinational companies, it said, may benefit the most.
APIs that update themselves
YC’s final example was Self-Maintaining APIs. It said API updates still depend heavily on manual work, and developers often suffer service interruptions when versions change.
YC wants each API to have its own AI agent. It gave Stripe as an example: when an API changes, an agent could scan customer codebases, make changes, submit pull requests and potentially complete the upgrade path end to end. The aim is to turn API maintenance into an automated process.
Across the Fall 2026 RFS, YC is clearly broadening its startup agenda beyond software alone, tying AI and crypto to defense, education, physical industry, aging societies and financial infrastructure.