AI Personalized Medicine
YC partner backing this: Ankit Gupta.
Doctors today guess. They prescribe drugs designed for the population average and hope they work for you. AI can read your genome, your bloodwork, your wearables, and your medical history at once — and recommend the treatment that actually fits.
YC wants companies turning that capability into real products: AI second opinions, personalized drug regimens, dynamic dosing, condition-specific care plans, longevity protocols.
Why YC wants this
Personalized medicine has been a research promise for 20 years. Foundation models trained on biomedical data finally make it shippable. The window to be the first consumer-grade product in this space is open right now.
Where the Money Is Direct-to-consumer health spend is enormous. The user pays $200/month for a personalized care plan because their insurance won't cover what their data actually says they need. You bypass the insurance maze entirely.
Real wedges to start with
- Personalized supplement and drug stacks based on labs and DNA
- Chronic condition management (diabetes, autoimmune, mental health) with AI-tuned protocols
- Pre-diagnosis triage that interprets symptoms + labs and routes to the right specialist
- AI second opinions on complex diagnoses, reading imaging and notes
- Longevity coaching that adapts protocols weekly based on biomarkers
Healthcare is regulated. You don't need to ship a diagnosis — you ship a personalized plan, education, or coordination layer and partner with licensed providers for the medical part. Most YC health companies start exactly this way.
Counter-Swarm Defense Stack
Drone warfare changed in the last 24 months. Every modern conflict now involves cheap autonomous drones, often launched in coordinated swarms. Existing air defense systems — built for jets and missiles — can't shoot them down economically.
YC wants the company that builds the counter-swarm layer: detection, tracking, classification, and interception across the full sensor stack.
Why this is now
Anduril went through YC. Today it's worth roughly $14 billion. The U.S. and allied governments are actively writing checks for defense tech that didn't exist five years ago. The bar to enter has dropped from "ex-Lockheed engineer" to "founder who can ship hardware + AI."
The Real Opportunity You don't need to build the whole stack. Pick one slice — passive RF detection, low-cost interceptors, edge-AI classifiers, sensor fusion software — and become the layer that the bigger primes have to license or acquire.
Wedges to consider
- Detection: low-cost radar + acoustic + RF arrays
- Sensor fusion: stitching multi-modal feeds into one threat picture
- Interception: kinetic, net-based, electronic warfare, or directed energy
- Software: AI classifiers running at the edge on cheap compute
- Logistics: rapid manufacturing and field repair of attritable systems
If you're not a defense insider, your edge is software speed. Most defense primes ship features on 5-year cycles. A YC-backed team can ship a usable counter-swarm classifier in 90 days. That gap is the entire business.
AI for Low-Pesticide Agriculture
YC partner backing this: Garry Tan, the CEO of YC.
When the CEO writes the RFS himself, that's the strongest signal you'll see. Garry wants a company that uses AI vision and robotics to cut pesticide use by 80–90% while keeping crop yields the same — or higher.
The play: cameras and AI on tractors, drones, or rovers identify exactly which plants need treatment, and treat only those. No more spraying entire fields with chemicals.
Why this matters
Pesticides are a $70+ billion global market, and farmers want to use less of them. Regulation in Europe, California, and parts of Asia is squeezing the chemical incumbents. AI precision agriculture is how you replace them.
The Founder Profile YC Wants You don't need to be a farmer. You need to ship hardware that runs in dirty conditions, partner with one mid-sized farm operator for the pilot, and prove the per-acre savings beat the system cost. That's the whole pitch.
Real angles
- Vision-based weed identification with targeted spraying or laser zapping
- Disease detection before it spreads, using drone imagery
- AI scouting reports for farmers — what's happening in each field, daily
- Soil and irrigation optimization with sensor + AI loop
- Bio-pesticide delivery robots that replace chemical spray entirely
Don't try to sell direct-to-farmer at first. Sell to the agronomy companies that already have farmer relationships. They become your distribution. You become indispensable in 18 months.
Company Brain
Every company today is drowning in unstructured knowledge. Slack threads, Notion docs, Google Drive, Linear tickets, customer calls, code repos, internal wikis — none of it talks to each other.
YC wants the AI that reads all of it and becomes the institutional memory. Ask it anything — "why did we kill that feature in 2024?", "what's the contract terms with vendor X?", "who knows about onboarding flow?" — and it answers like a senior employee who's been there for a decade.
Why this is the biggest of the five
Search inside companies is broken. Glean and a handful of others got partway there, but the category is still wide open. Whoever nails the agentic version — that not only answers but takes action — owns the operating system layer for every modern company.
The Defensible Moat Once a company connects all its data to your product, switching costs go through the roof. The AI gets smarter the longer it lives at that company. You become unrippable in 12 months. This is the kind of business that compounds.
How to wedge in
- Pick one role first: AI for new hires, AI for support engineers, AI for sales reps
- Pick one company size: 50–500 employees is the sweet spot — big enough to have the problem, small enough to buy fast
- Pick one trigger: someone asks a question in Slack, your bot answers using all the context
- Charge per seat or per query — both work; per-seat is easier to start
This space is going to get crowded. The winners will be the teams that ship the best memory architecture and the best agent UX, not the ones with the most enterprise sales reps. Stay lean. Stay product-first.
Beyond the official five
YC's RFS is the public list. Recent batches show patterns of what they actually fund alongside it. These 15 themes match the kinds of companies getting checks from YC right now.
06
AI coding agents for non-developers
Tools that let an ops manager or marketer ship internal apps without writing code.
07
Voice agents for inbound and outbound calls
Sales, support, scheduling — replacing the BPO industry one workflow at a time.
08
AI for legal and compliance
Contract review, redlining, discovery, regulatory monitoring at 1/10th the cost.
09
Robotics for the trades
Plumbing, HVAC, electrical, painting — labor shortages mean any robot helper sells fast.
10
Climate adaptation infrastructure
Wildfire detection, flood prediction, grid hardening — paid customers already exist.
11
AI for drug discovery
Lead generation, target validation, clinical trial design — biotech is finally programmable.
12
New space economy
Earth observation, in-orbit services, satellite swarms — launch costs collapsed; payloads haven't caught up.
13
AI tutors and adaptive learning
K–12, professional certifications, language — personalized instruction at $20/month.
The rest of the list
14
Manufacturing software with AI agents
Quality control, scheduling, predictive maintenance for mid-sized U.S. factories.
15
Personal AI for households
Bills, taxes, scheduling, kids' admin — the chief-of-staff every family wishes they had.
16
AI for civic and government services
Permits, benefits eligibility, 311 calls — public-sector software is decades behind.
17
Construction automation
Bidding, scheduling, on-site monitoring — the $13T construction industry is starving for software.
18
AI cybersecurity agents
Autonomous threat detection and response — every CISO is buying right now.
19
Energy grid software
Battery dispatch, virtual power plants, EV charging coordination — utilities pay real money.
20
Translation and multilingual AI
Voice + video translation that preserves tone and context, for global commerce and content.
How to Use This List Don't pick the idea that sounds smartest. Pick the one where you have an unfair edge — a relationship, a skill, a frustration you've lived through. YC funds founders who explain why they are the one to build it, not idea optimizers.
The weekend MVP stack
Here's the part nobody tells you: you don't need to know how to code to ship a YC-grade prototype anymore. The AI dev tools released in the last 18 months are good enough that a non-technical founder can build a working demo in a weekend.
This is the exact stack I use and recommend.
Claude Code — your senior engineer
Anthropic's terminal-based agent. You describe what you want in plain English. It writes the code, runs the tests, fixes the bugs, and commits to GitHub. Free with a Claude Pro subscription. Best for: backend, scripts, agents, anything that requires real reasoning.
Open a fresh terminal, run claude, then say "build me a Next.js app that does X." It plans, scaffolds, and ships. The first 60 minutes will feel like cheating.
Cursor — the IDE that writes with you
Forked VS Code with AI built into every keystroke. Highlight any function and ask it to refactor, explain, or extend. Best for: when you want to read along as the AI writes, or when you already know enough code to steer.
v0 by Vercel — UI in seconds
Type "design a landing page for a YC application" and v0 generates production-ready React + Tailwind. Drop it into your repo. Best for: marketing pages, dashboards, polished UI without hiring a designer.
Replit Agent — zero-setup deploys
The whole loop — code, database, hosting — inside one browser tab. Tell the agent what to build; it ships a live URL. Best for: getting a clickable prototype in front of users on day one.
Lovable — full-stack AI builder
The "I want a SaaS by Sunday" tool. You describe the app, it builds the full thing — auth, database, frontend, payments. Best for: validating an idea with paying users before you write a line of code yourself.
The Real Workflow Use Lovable or v0 to ship the marketing site and waitlist in 30 minutes. Use Claude Code to build the actual product in 2–5 days. Use Cursor to maintain it. That's the entire YC-application-ready stack — and it's all under $100/month.
From this guide to the YC application
Reading this guide is the easy part. Below is the exact 7-step path to apply with a real prototype before the deadline.
1
Pick one idea (today)
From the 5 official ideas or the 15 bonus list. The one where you have the strongest "why me" story wins.
2
Write your one-paragraph pitch (today)
Problem, solution, who pays, why now. If you can't fit it in 4 sentences, the idea isn't sharp enough yet.
3
Talk to 5 potential users (this week)
Cold outreach on LinkedIn, X, or in real life. Just listen. The product builds itself once you've heard the same complaint 5 times.
4
Ship a prototype with Claude Code or Lovable (weekend)
Doesn't need to be production-ready. It needs to be clickable. Put it in front of the users from step 3.
5
Get one paying customer (next 2 weeks)
Even $50/month. Even an LOI. Revenue in any form is the single biggest signal in a YC application.
6
Record the YC video (1 hour)
One minute. Founders on camera. No slides. Show the product. Say the numbers. That's the entire script.
7
Submit before the deadline
Apply at ycombinator.com/apply. Read the full RFS at ycombinator.com/rfs. Late applications still get reviewed but lose priority — submit early.
"The best YC applications I've seen weren't from the smartest founders. They were from the ones who shipped fastest." — Common YC partner advice