AI Startups Ideas by YC, Summer 25
What Top VC’s Look For
2025 is shaping up to be the year of AI-native startups.
And Y Combinator just dropped another confirmation of it: their latest Requests for Startups (RFS) — a list of startup ideas they want to fund.
This isn’t trend-chasing. These are invitations. If you’re building with AI (or thinking about it), this is your signal.
So in today’s post, we’re diving into what YC’s RFS says about where startups are headed:
– What is “full-stack AI,” and why is it important – Why YC wants more designers, scientists & teachers to become founders – The rise of AI agents, internal bots, and voice assistants – Sectors begging for reinvention: healthcare, security, education, finance – And some wild startup ideas hiding between the lines
Let’s unpack it all — fast, clear, and idea-packed.
This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.
# The “Full-Stack AI” Mindset
Most founders build AI tools and try to sell them to slow-moving incumbents. YC’s take? Don’t sell to the dinosaurs — make them extinct.
“You could build an AI agent and sell it to law firms... Or, you could start your own law firm, staff it with AI agents, and compete with the existing law firms. That, my friends, is going full-stack.” — Jared Friedman, YC
This is the kind of reframing that hits hard. Don’t build for the system — rebuild the system. A school run by AI tutors. A clinic run by AI agents. A logistics company with zero manual coordination.
And it’s already happening.
🏛️ Garfield AI
Garfield AI is a fully AI-powered law firm in the UK. It helps clients send debt recovery letters for £2 and file lawsuits for £50 — no lawyers involved, just an AI interface guiding users through the entire legal process. Even more impressive? Garfield received official approval from the UK’s legal services regulator (SRA) — a huge signal that AI-native institutions are not only possible, but already here.
YC calls it full-stack AI. It’s not just software — it’s owning the whole vertical, rebuilt with intelligence baked in.
# New Roles, New Founders
If full-stack AI is about rethinking industries, the next question is: who should build them?
YC has a clear answer: designers, teachers, scientists — this is your moment.
“Designers already have so many of the skills needed to be great – strong user empathy, a focus on solving problems, a high bar for quality, and taste.” — Aaron Epstein, YC
The rise of AI doesn’t mean your job is getting replaced. It means your insider knowledge is now 100x more powerful — if you become the founder.
Design is more than how things look — it’s how people trust and engage with products. That’s why companies like Airbnb and Stripe couldn’t have worked without first-class design.
Same goes for scientists and educators: AI is now breaking open bottlenecks in drug discovery, simulations, and personalized learning. If you've spent years in a field, you know the pain points — and now you have the tools to solve them at scale.
So don’t wait to be hired. YC’s saying it loud: you should be the one building what’s next.
And it’s also already happening:
Alice. Tech is an edtech startup from Copenhagen founded in 2024. It just raised $4.8M to build an AI-powered personalized learning platform. The twist? Founder Kim Rants isn’t a technical founder. Kim Rants is an educator using AI to turn dense materials into adaptive content — flashcards, explanations, and quizzes — tailored to each student’s learning style.
That’s the formula: domain insight + AI leverage. You bring the expertise. AI brings the scale. Normalize background diversity.
Thanks for reading! Feel free to share it.
# AI Agents Everywhere
So, we’ve got new kinds of founders. And they’re not just building dashboards. They’re building agents.
“I want to see a system that truly does the tasks a human personal assistant or chief of staff might otherwise do.” — Tom Blomfield, YC
We’re talking internal bots that handle your repetitive workflows. Voice agents that pick up calls, answer emails, and triage your calendar. Assistants that don’t just remind you about a task — they get it done.
This isn’t sci-fi. It’s already shipping.
📥 YC built internal agents to automate their work — everything from reviewing term sheets to handling accounting ops. 🗣️ Founders are using tools like Vapi and Retell to build voice interfaces so good, you’d swear they’re human. ✉️ Some are even building inbox agents that help you hit “Inbox Zero” before you arrive at the office.
🧩 With Modular Command Protocols (MCP), Cursor can now access 80% of Figma’s functionality programmatically.
It means design tasks like asset editing, layout changes, or UI tweaks can now be automated, streamlined, and integrated right into your dev environment. Check more examples with MCP in our article.
The shift is clear: from tools to teammates. The new generation of AI doesn’t just assist — it acts.
# Sectors Ripe for Reinvention
Once you’ve got agents that act, the next question is: where should they act first?
YC’s answer: the big, broken sectors where AI can create immediate leverage. Healthcare. Finance. Education. Home security. All massive. All inefficient. All ready.
🩺 Healthcare
Healthcare isn’t just hospitals — it’s everyday decisions. What you eat. How do you live? And most people have no clue how their daily habits stack up.
That’s where AI can quietly do something huge: make healthy choices effortless.
Cal.ai is going viral for a simple reason: it makes nutrition frictionless. Take a photo of your food, and it instantly tells you the calories, macros, and nutrition profile — powered by computer vision and smart reasoning.
No scanning barcodes. No logging ingredients. Just snap and go.
And why is it blowing up? Because it makes something hard and annoying feel effortless. It turns intention into action with zero overhead — and that’s the recipe for virality in healthtech.
💸 Personal finance
It’s still a mess — biased advice, expensive planners, Google searches as default.
“Most humans are not rational about their finances… With LLMs, there’s a unique opportunity to give everyone personalized, unbiased advice.” — Gustaf Alströmer, YC
💬 Cleo
Cleo — an AI chatbot that helps you budget, save, and spend smarter with zero judgment and a ton of personality. Finance that talks like a human.
🎓 Education
Education is stuck in 1999 — one-size-fits-all courses and slow systems.
“It’s impossible to believe that education is still going to be done in the same way it is today.” — Tom Blomfield, YC
🧑🏫 Flint
Flint (YC S23) is building an AI-powered platform that helps schools deliver personalized learning at scale.
Teachers can generate worksheets, quizzes, and lesson plans in seconds, tailored to students’ needs and aligned with curriculum standards. No more endless prep. Just better, faster teaching.
Flint’s not replacing teachers — it’s giving them a superpower. And for students, that means learning that adapts to how they think, not the other way around.
🏠 Home security
It is still built on tech from the ‘90s — motion sensors, blinking lights, maybe a loud beep.
“The biggest players are legacy companies whose features haven’t changed in decades… Imagine how big the opportunity is for the company that uses AI to make people feel actually safe.” — Tyler Bosmeny, YC
Deep Sentinel pairs AI cameras with real-time threat detection and even live guards who speak through loudspeakers to stop break-ins before they happen.
AI as Infrastructure
Once the products are in place, it’s time to look under the hood — at the infrastructure powering it all.
“This infrastructure will need to access every other piece of software… manage permissions, and send sensitive data securely to the latest generation of LLMs.” — Pete Koomen, YC
Agent builders, memory systems, permission layers — these are the new rails for AI-native companies.
“We’re interested in funding people building software tools to help others make robots.” — Diana Hu, YC
A great example of this is RoboFlow
Roboflow don’t build robots themselves, but provide essential tools (like data management and model training for computer vision) that help robotics developers add vision to their systems.
They went through YC in 2020 and are now powering thousands of real-world robotics and AI applications.
If AI is the engine, this is everything that makes it run.
Was it worth it? Share it with friends!
# Conclusion
We’re not just seeing better tools — we’re seeing entire industries up for grabs. Founders aren’t just building apps. They’re building AI-native institutions. Designers are becoming CEOs. Teachers are becoming product thinkers. Scientists are becoming builders.
We’re at the point where AI isn’t the product — it’s the foundation. The next great companies won’t be AI-enabled. They’ll be AI-native — full-stack, domain-informed, and infrastructure-powered.
And 2025 might be the year it all tips.
If any of this made your brain light up — good.
YC’s Requests for Startups aren’t just signals. They’re open invitations. You don’t need permission. You don’t need credentials. You need curiosity, urgency, and the guts to build.
The best startup ideas don’t sound crazy anymore — they sound inevitable.
So go browse the full list. Pick the one that won’t leave your head. And if you’ve already got that itch?
It’s probably time to scratch it.
This article was first published in the Creators AI newsletter. View the original edition.
