The most unusual RFS in YC's history
Y Combinator's Fall 2026 Requests for Startups is unusual. Not because of the number of AI ideas — that's expected. Because one of the thirteen requests is signed by Daniel Driscoll, the sitting US Secretary of the Army. That has never happened before in RFS history.
A cabinet-level federal official published a founding-team-level ask in the same document as a request for AI tutors, multiplayer agents, and offshore compute flotillas. The signal isn't just that YC and the Army are now aligned. It's that the boundaries between commercial software startups, physical-world builders, and federal government procurement have collapsed further than most founders realize.
That's just one signal. Reading the full document, four dominant themes emerge that together map out where YC thinks value is being created next — and what founders trying to decide what to build should pay attention to.
Why RFS documents matter
YC's Requests for Startups are the closest thing to reading their strategic mind. They don't publish theses. They publish what they want built. Each request is written by a partner or high-signal alum with specific opinions about a specific problem.
Reading the RFS is like reading Berkshire Hathaway's annual letter — not necessarily what they'll bet on directly, but what they think matters. Every founder deciding what to build in the next 12 months should read this document. It's free market intelligence from the most active early-stage investor in the world.
Here's what the current one says.
Theme 1: AI moves into the physical world
The largest single theme in the RFS. Five of thirteen requests are explicitly about AI extending into physical space.
The Future of American Defense by Daniel Driscoll, Secretary of the Army — asking for low-cost interceptors, next-gen sensors, drones, resilient logistics, and advanced manufacturing that can operate in extreme climates. AI for the Aging Population by Max Kolysh — asking for voice interfaces, monitoring systems, robotics for physical assistance, and software for family caregivers to serve the 1-in-5 Americans who will be over 65 by 2030. New Operating Systems for the Physical World by Charlie Warren — asking for operating systems that can route work between AI agents, robots, and humans working side-by-side in construction, maintenance, and fleet operations. Data for the Real World by Austin Tindle and Diana Hu — asking for companies collecting physical-world data through robots, sensors, and autonomous instruments in energy, agriculture, logistics, and construction. Compute at Sea by Francois Chaubard — asking for founders building offshore data center flotillas because land-based data centers are running out of land, power, and community support.
The pattern is consistent. AI is moving from screens to sensors, robots, and physical infrastructure. YC is explicitly telling founders that the interesting work is where AI meets atoms, not where AI meets more screens.
Theme 2: AI as a shared team member, not a personal assistant
Three requests focus on turning AI from a solo tool into infrastructure for teams and organizations.
Multiplayer AI by Aaron Epstein — arguing that the single-player AI moment (ChatGPT-style solo chats) is ending, and that agents running multi-day tasks need to be shared, redirectable, and hand-offable across humans and teams. A Cloud for Small Software by Pete Koomen — asking for infrastructure to easily deploy and share the "small software" that agents can now build for individual teams and workflows, without the complexity of AWS or Azure. Self-Maintaining APIs by Harsha Gaddipati — arguing that agentic coding tools have created a new possibility: when Stripe or any API vendor ships a breaking change, an agent should scan customer codebases, identify affected usages, and open a PR with the fix.
The theme underneath: AI infrastructure for the AI-native era. Not tools that help humans use AI. Tools that assume AI is doing the work, and build the substrate for humans to collaborate with agents at scale.
A cabinet-level federal official published a founding-team-level ask in the same document as a request for AI tutors and offshore compute flotillas. That has never happened before.
Theme 3: The trust layer of the internet is broken
Two requests explicitly address a category that barely existed 24 months ago.
Proving You're Human by Max Kolysh — pointing to the finance worker who wired $25 million on a deepfake video call, and arguing that "every trust signal we have was built for a world where faking a human was expensive, and that world is gone." AI-Native Compliance Infrastructure by Daivik Goel — pointing to the regulatory patchwork that varies by state, and the fragmented compliance tools that force companies to hire chief compliance officers and specialized headcount.
The trust and compliance layers of the internet were built assuming humans on both sides of transactions. AI collapses that assumption. YC is calling out that entire new categories need to be built to restore the assumptions that commerce, communication, and regulation rely on.
Theme 4: Consumer AI is coming back
Three requests argue that despite three years of AI hype producing almost no new consumer icons, the moment is now approaching.
AI-Powered Consumer Products for 1 Billion People by Raphael Schaad — arguing that intelligence just got good enough to treat agents like people, and cost is falling 10x per year, meaning "whoever builds now, owns it." The Primer by Andrew Miklas — asking for a Stephenson-inspired adaptive tutor that teaches young children to read, write, and reason at the quality of a devoted private tutor. The Best Time to Build in Crypto by Nemil Dalal — arguing that bear markets are when real projects build, and that crypto rails will underpin consumer applications from capital raising to payments.
The pattern: after three years of enterprise AI dominating, YC is calling attention to consumer as the next wave. The economics have finally moved to where mass-market AI products become viable, and YC wants founders positioned for that shift now, not two years from now.
The signal that matters most
The Army request is not just symbolically important. It signals that federal government procurement has fundamentally opened to startups in a way that was unthinkable five years ago.
Anduril proved the model — a defense-tech startup could become a multi-billion-dollar business by shipping modular hardware and software into government contracts. Palantir showed the scale. Shield AI, Epirus, and dozens of others have built genuinely large defense-tech companies in the past decade.
Now the sitting Secretary of the Army is publishing directly in YC's RFS. This is a permission structure change more than a category change. It signals that the highest levels of the US federal government are actively courting startup founders — not just Silicon Valley's biggest names, but early-stage companies with the right ideas.
Founders considering defense-tech in 2026 have a market environment that would have been unimaginable in 2020. And the same permission structure is arguably opening up in healthcare (aging population request), infrastructure (compute at sea), and public education (Primer). Government-adjacent categories are becoming a viable path for early-stage startups in ways they weren't a decade ago.
What to do about it
Three concrete actions:
If you're a founder deciding what to build — take the physical-world theme seriously. Consumer software AI is saturated. Enterprise AI copilots are saturated. Physical-world applications — defense, aging care, construction, infrastructure — are where competition is thinnest and demand is strongest. The permission structure has never been more favorable.
If you're an investor evaluating theses — the trust infrastructure category is genuinely under-invested relative to its likely importance. Deepfake fraud is already at real scale. AI compliance is being demanded by every enterprise. The companies that build the trust layer for the AI era will be foundational.
If you're a builder thinking about your next role — read the RFS not as a job board but as a career signal. The categories YC is calling attention to are the categories that will get heavily funded and where operators with early exposure will be most valuable in 18-24 months. Position accordingly.
The takeaway
YC's Fall 2026 RFS maps out where they think the next wave of company formation will happen. AI moving into the physical world. Multiplayer agents replacing solo AI tools. Trust infrastructure being rebuilt for an AI-native era. Consumer AI becoming economically viable. And a permission structure change that lets defense, healthcare, and infrastructure become viable early-stage categories in ways they weren't a decade ago.
This document is worth reading in full. It's the closest thing to a strategic map of what YC — the most influential early-stage investor in the world — thinks matters right now. Founders who build for these categories in the next 12 months are building with wind at their back.
Want to understand the financial mechanics behind AI-native businesses across all these categories — pricing models, unit economics, and the strategic shifts reshaping software? Try the AI Business Models topic pack on Gargiulo — scenarios covering usage pricing, outcome pricing, and the emerging economics of the AI-native economy. Sterling has notes.
Source: Y Combinator Fall 2026 Requests for Startups.