YC Requests for Startups — Fall 2026
- URL: https://www.ycombinator.com/rfs
- Date Saved: 2026-07-29
- Source: Y Combinator website
- Tags: business, ai-engineering
Summary
YC’s tradition of sharing startup ideas they want to fund. These are “extra validation” — you don’t need to work on these to apply. Theme this batch: AI moving into the physical world. Includes first-ever RFS from the sitting U.S. Secretary of the Army.
The Ideas (14 total)
1. AI Tutor (The Diamond Age Primer)
- Inspired by Neal Stephenson’s “A Young Lady’s Illustrated Primer”
- Start with: adaptive teaching of reading, writing, arithmetic for young children
- Quality of a private tutor at consumer scale
- Long-term vision: AI that grows with the child, teaches thinking/reasoning/wisdom
- Not replacing teachers — supplementing them
2. Defense Tech for the U.S. Army
- Written by the Secretary of the Army
- Low-cost interceptors / lower cost-per-kill
- Next-gen sensors, software, payloads for open system architecture
- Drones, resilient logistics, advanced manufacturing
- Must survive extreme climates (Arctic to archipelago, space to subterranean)
3. Small Software Cloud
- “Small software” = purpose-built tools for one user or small team
- Easy to build with agents, still hard to deploy and share
- Need a cloud designed for small software (not AWS/Azure complexity)
- Challenges: auth/permissions, security for arbitrary code, customization
- Goal: sharing software should be as easy as sharing a Google Doc
4. Multiplayer AI Agents
- AI is still single-player (private chat threads)
- Work tools won by going multiplayer (Google Docs, Figma)
- Agents now run tasks taking hours/days/weeks — needs team collaboration
- Anyone should be able to drop into a live agent session, redirect, hand off
- Applies to: engineering, sales, support, legal, marketing — any team work
5. Offshore Compute (Data Centers on the Ocean)
- AI running out of compute, data centers out of electricity/land
- New data centers take years + local gov opposition
- Ocean = 70% of Earth surface, abundant sunlight, no permitting, natural heat sink
- Concept: “compute flotillas” — modular vessels as one global cloud
6. AI-Powered Consumer Products for 1 Billion People
- Every platform shift mints consumer giants (web → Google/Airbnb, mobile → Instagram/DoorDash)
- AI is biggest shift yet, but only ChatGPT is on home screens 3 years in
- Now: intelligence good enough to treat agent like a person
- Costs falling 10x/year — consumer moment landing very soon
- Domains: productivity, transport, learning, health, finance, play, social
7. AI for Aging Population / Elder Care
- By 2030, 1 in 5 Americans over 65
- Millions of unfilled caregiving jobs, 53M family members doing it unpaid
- Almost no tech built for older people (even Alexa frustrates seniors)
- AI enables: voice interfaces with real conversations, safety monitoring, home robotics, care coordination software
8. New Operating Systems for the Physical World
- 80% of global workforce doesn’t sit at a desk
- Software for construction/maintenance/fleet hasn’t changed in 20+ years
- Now 3 types of workers: AI agents, robots, humans with wearables
- Need new OS to manage all three together
- Opportunity much bigger than existing software — industries spend 10-100x more on labor than software
- Competitive advantage: you’ll record all work as it actually happens
9. Crypto (Bear Market = Best Time to Build)
- YC invested in 100+ crypto startups, expects that to grow a lot
- Expect every YC startup to eventually use crypto rails
- Key trends: stablecoins adopted by major financial institutions, tokenized stocks, agents using crypto as financial rails
- Regulatory clarity finally here
- Excited about: capital raising, stablecoins, agentic commerce, trading, institutional products, private blockchains
- Bear markets attract better founders (building > speculation)
10. Dense Physical-World Data Collection
- AI is superhuman for code/language/images but physical world data is sparse
- Improving foundation models + plummeting sensor costs = dense data now feasible
- Examples: Gecko Robotics (robots for hard-to-reach data), Sorcerer (autonomous weather balloons)
- Biggest industries (energy, agriculture, logistics, construction) rely on limited data
- More data → precise models → control systems (steering hurricanes, reversing desertification)
11. Rebuilding Internet Trust Layer (Anti-Deepfake)
- $25M stolen via deepfake video call — all “colleagues” were fakes
- Voice clones + fake video getting cheap and ultra-realistic
- Need: verified human on other end of any call/message/transaction
- Ideally without sacrificing privacy
- Use cases: Twitter without bots, dating apps with real people, verified reviews
- Whoever builds this becomes the layer every bank/app/video call checks
12. AI-Native Compliance Infrastructure
- Financial compliance still uses spreadsheets, siloed tools, expensive headcount
- Complexity compounds as businesses expand to new markets
- Most compliance work (monitoring, flagging, reporting, audit trails) = AI tasks
- Pain point: state-by-state licensing, renewals, audits, regulatory patchwork
- Want: consolidated tools, less specialized headcount, real-time visibility across jurisdictions
13. API Communication Agents (Auto-Apply Breaking Changes)
- API communication is broken — breaking changes ship with little warning
- 30% of AWS service downtime was from unnoticed external API/package changes
- Developers now give codebase access to external tools (Claude Code, Devin, etc.)
- Missing piece: application layer connecting API providers to customer codebases
- When Stripe ships a breaking change → agent scans customer code → opens PR with fix
- Could be per-provider agents or neutral third-party (like Dependabot but for APIs)
Key Themes
- AI in physical world (defense, elder care, construction, data collection)
- AI going multiplayer/collaborative
- Infrastructure for the AI era (compute, trust, compliance, small software)
- Consumer AI moment is imminent
- Crypto optimism in bear market