Y Combinator's Requests for Startups (RFS) is their traditional way of sharing ideas they hope founders will tackle. These ideas represent only a fraction of what YC actually funds — if one of them excites you, that's extra motivation to dive in. But you don't have to focus on any of these ideas to apply to YC.

"If one of these excites you, that itself is additional conviction to dive in. But you don't have to focus on any of these ideas to apply to YC."

Summer 2025: The Year of AI Agents

2025 is shaping up to be the year of AI agents. YC has compiled a list of AI agent startup ideas they find especially promising. Some reflect trends that have already taken hold; others point toward where things are headed.

1. The Era of Full-Stack AI Companies

YC wants to support more founders building full-stack AI companies.

"Wondering what a full-stack AI company is? Imagine that LLMs can now automate many legal tasks. Most people would build an AI agent and sell it to law firms. But you could also just start your own AI-powered law firm and compete directly with existing ones. That's full-stack."

In other words, instead of selling solutions to existing companies, you enter the industry yourself and replace those companies.

"Don't sell to the dinosaurs — make them extinct."

This approach holds particular promise in industries that are slow to change.

2. The Era of the Designer Founder

Over the next decade, advances in coding tools will make it faster and easier to build and ship products — which means design will matter more than ever.

"Designers worry about AI taking their jobs, but the real opportunity is using AI to build products yourself and start a company."

Designers already have the skills that matter most for founding: user empathy, problem solving, quality standards, and taste.

"YC's flagship companies like Airbnb and Stripe succeeded in large part because of exceptional design. Without it, you wouldn't stay in a stranger's home or trust Stripe with billions of dollars in credit card transactions."

YC is eager to support founding teams where design is central.

3. Voice AI Agents and the Future of the Phone Call

The phone call has barely changed as a business communication medium in the past hundred years.

"Long hold times, confusing voice menus, prompts to press star or pound — in the end, we always want to talk to a real person. Because the alternatives were so bad."

But the latest voice models and conversational LLMs have produced voice AI bots that are indistinguishable from humans.

"Talking to a voice AI bot feels like experiencing the future — like riding in a self-driving car for the first time. It just works."

More than a trillion business-to-customer calls happen every year. This is a market ripe for AI-driven reinvention.

4. Innovation in Scientific Software

Software for chemistry, biology, materials science, and operations research has gone largely unchanged for decades.

"These fields still rely on standard methods and PhD-level talent to solve complex problems."

But advances in test-time compute and AI are opening the door for new startups in drug discovery, chemical process optimization, metals and mining, power grid optimization, and more.

"We want to see startups using AI to reinvent the way physical things are made — faster and more efficiently."

5. AI Personal Assistant: From To-Do List to Done List

Productivity apps have evolved for decades, yet:

"Email piles up, calendars fill up, and tasks keep getting postponed."

Existing tools tell you what needs to be done but don't actually do it.

"Now, with advances in LLMs, we can move from the 'to-do list' to the 'done list.'"

YC wants startups building AI personal assistants that genuinely do the work.

"Imagine an AI that fully understands your work, your routines, your communication history, and your personal preferences — and actually acts on your behalf."

For example:

• Drafting and automatically sending email replies
• Automatically scheduling meetings
• Handling repetitive tasks autonomously

"This isn't just message filtering or auto-filling a calendar. We need a system that does what a real human assistant or chief of staff does."

6. Automating US Healthcare Administration

The US healthcare system accounts for 17% of GDP — roughly $4 trillion per year.

"More than a third of that — over $1 trillion — is spent on administrative work."

The reasons: lack of interoperability between systems, no APIs, manual data transfer, and more.

"Over the past two years, infrastructure startups have emerged that extract data from PDFs and other sources, normalize it, and make it easy to enter into other systems."

With advances in LLMs:

"Many administrative tasks can now be fully automated. This is a chance to make the US healthcare system dramatically more efficient."

7. AI-Powered Personalized Learning Tools

The dream of using computers to help people learn dates back to the 1940s.

"Vannevar Bush's Memex, JCR Licklider's 'man-computer symbiosis,' Alan Kay's Dynabook..."

But truly personalized learning experiences have never been realized.

"Most online education delivers the same content to everyone."

Now, with AI's latest reasoning capabilities and multimodal models (text, voice, images, 3D, and more):

"Complex topics can be broken down step by step, with concepts visualized through animation, 3D, voice, and more."

"We've already seen the impact a single great teacher can have — Grant Sanderson (3blue1brown) on YouTube makes complex concepts accessible to millions. Now imagine an AI tutor that delivers that quality of explanation for every subject, personalized to every learner."

8. Reinventing the Education Industry

Education is one of the largest and most important industries in the world, yet:

"It has barely changed in 100 years. One billion people work in education, and 1.5 billion receive it every year."

The rise of AI, especially LLMs, signals major changes ahead in:

"Access to education, personalized instruction, and freeing teachers and learners from repetitive work."

"We're only at the beginning of exploring what AI can do in education — personalized learning tools for students, grading platforms for teachers — it's early days."

The business model is a significant challenge, but:

"It's hard to believe education will look exactly the same in 10 or 20 years."

9. AI-Powered Home Security

Consumers spend $20 billion per year on home security.

"The major players have seen almost no feature change in decades."

Meanwhile, commercial security is being rapidly reinvented by AI:

• Facial recognition access control
• Video-based anomaly detection
• AI guards deterring crime through speakers

These are already commercially available.

"Ring was acquired for a billion dollars with a video doorbell and some sensors. The opportunity for a company that makes people feel genuinely safe at home using AI is enormous."

10. Internal Agent Builders: Automation for Every Employee

Soon, every company will have:

"Every employee building their own agents to automate their repetitive work."

YC is looking for founders to build the internal agent builder infrastructure.

"These tools need access to all the software I use every day, with permissions management and secure handling of sensitive data."

YC is already using tools like this internally:

"We save time on things like glossary review and repetitive accounting tasks — which lets us focus more on what only humans can do well."

11. AI Research Labs

YC wants to support more AI research labs.

"Something many people don't know: YC was OpenAI's first investor. In fact, OpenAI started as an internal research lab called YC Research."

Like OpenAI:

"The concept of an independent AI research lab was pioneered there, and we watched that journey up close."

"Many people think YC requires you to ship a product in three months — but we're just as willing to support deep, open-ended research that may take years, like OpenAI."

12. Voice-Based Email Assistant

"What if, during your 20-minute commute every morning, you could sort through your email, draft replies, and delegate scheduling — so that by the time you get to the office, you're already at inbox zero?"

The latest voice agents — Vapi, Retell, ChatGPT Voice Mode — are already remarkably capable.

"After experiencing nerve damage myself, I realized how much we depend on keyboards. Email might be the perfect starting point for building a universal assistant."

"If you have access to my inbox, you know my friends, my plans, my writing style — everything."

13. AI-Powered Personal Finance, Investment, and Tax Advice

"Most people are not rational about their finances. But to get through life, you have to decide how much to save for the future, where to invest, and how to manage debt and taxes."

Until now, the options were:

• Asking a friend
• Googling
• Hiring a financial advisor
• Calling your bank

"These approaches are biased and don't reflect your full financial situation and goals. And great advisors are extremely expensive."

With LLMs, it's possible to build software that:

"Gives everyone nearly free, fully personalized financial, investment, and tax advice."

By connecting via API to your complete financial picture:

"Completely personal, unbiased advice becomes possible."

Key Themes Summary

• AI agents
• Full-stack AI companies
• Design-driven founding
• Voice AI / call innovation
• Scientific and engineering software innovation
• AI personal assistant
• Healthcare administration automation
• AI personalized learning
• Education industry innovation
• AI home security
• Internal agent builders
• AI research labs
• Voice-based email assistant
• AI personal finance advice

YC is looking for founders and teams who will change the future.

"If even one of these excites you, go for it. And even if none of these is your idea, YC is waiting for your next bold move!" 🚀