YC Partners Reveal: How to Strike Gold with AI Startup Ideas in 2025

In the AI startup wave, the biggest challenge for technical talent isn't technical capability—it's finding truly compelling startup ideas worth pursuing. Y Combinator partners Gary, Jared, Diana, and Harj reveal for the first time all the secrets they use in office hours, teaching you how to discover the next billion-dollar AI startup opportunity.

Why Can't Smart Technical People Find Good Ideas?

Many excellent AI engineers and technical experts face the same dilemma: they completely believe in AI technology's massive potential, know this is an excellent time to start a company, but just can't find startup ideas that truly excite them.

YC partners have observed that most failed startup ideas share a common trait: they're too lazy and superficial. These ideas typically come from:

• 📱 Hackathon Mentality: Seeing trending topics on X (formerly Twitter) and following along
• 🔥 Chasing Hot Trends: Blindly jumping on popular technology bandwagons
• 🏠 Staying in Comfort Zone: Only looking for opportunities in familiar territories

Core Problem: You need to get out of your house. Either dig deep into your internal expertise or aggressively explore the external world.

Method 1: Mining Internal Expertise - Are You Already an Expert?

Finding Your Unique Advantages

Many entrepreneurs already possess unique professional advantages, but they often don't realize it. YC partners shared several typical cases:

Case 1: Salient - AI Voice Agent

• Founder Background: Former Tesla Finance Operations team member
• Discovery Process: Discovered lease payment recovery process was entirely manual and inefficient while working at Tesla
• Solution: Developed AI agents for automated auto debt collection
• Current Status: Now serving multiple major banks

Case 2: Diode Computer - AI Circuit Board Assistant

• Founder Background: Both electrical engineers and software engineers
• Unique Advantage: Cross-skills in hardware and software are extremely rare in the industry
• Discovery Process: While building circuits at Apple and other companies, found hardware engineers lacked software tools like Git
• Solution: AI automation for component and datasheet verification

Case 3: Spur - AI QA Agent

• Founder Background: Former Figma engineer
• Discovery Process: Discovered frontend testing was extremely complex at Figma, with engineers spending enormous time writing and maintaining tests
• Solution: AI agents that automatically write test code

Mining Gold from Internship Experiences

Surprisingly, many of YC's billion-dollar companies can be directly traced back to founders' internship experiences.

Case: Data Curve

• Founder: 19-year-old college dropout
• Turning Point: Returned to previous internship experience at Cohere
• Discovery: Former boss needed synthetic and real data for large language models
• Achievement: Started last June, now generating mid-to-high seven figures annually

Key Insight: If you want good startup ideas, do internships at companies truly on the cutting edge of technology during college.

Method 2: Pursuing Life Mission - What Excites You for Life?

Sometimes the best startup ideas come from things you genuinely care about.

Case: Can of Soup Founder Gabriel was an early engineer at Substack, later lost confidence in his original idea. His former boss Chris gave him advice that changed everything:

This advice got Gabriel thinking about big ideas that truly excited him, eventually creating Can of Soup—an AI-powered entirely new social network.

Case: Happen Stance

• Founder: Former Apple AI researcher who sold his previous company
• Discovery: LinkedIn search still uses primitive text indexing, completely unintelligent
• Solution: Created intelligent search engine using LLMs and vector search
• Achievement: Can understand fuzzy queries and accurately find target contacts

Method 3: Getting Out of Comfort Zone - Becoming an Industry Detective

After mining your internal expertise, you need to actively explore the external world.

Leveraging Family and Friend Networks

Case: ESS Health - Dental AI Assistant

• Discovery Process: Founder worked a day with his mother (a dentist)
• Observation: Large amounts of insurance processing and pre-authorization work were routine procedures
• Solution: LLM-powered back office system for dentists
• Results: Performing exceptionally well

Why Are Family Networks So Important?

• You might send 1000 cold emails to strangers and get 0 responses
• But relatives you see at Thanksgiving will always give you some access
• Access is everything—you just need entry into an underserved place

Undercover Strategy: Getting Inside Jobs

The most extreme but effective method is working directly in the target industry.

Case: Medical Billing AI Company

• Strategy: Founder actually got a remote job as a medical biller
• Process: Built AI agents on his own MacBook Pro using Llama 3 to replace himself
• Legality: Using your own computer and open-source models is completely legal
• Learning: Deep understanding of industry operations from the inside

Another Case: Able Police

• Trigger Event: Founder's friend became a crime victim
• Deep Research: Participated in police ride-alongs, discovered police drowning in paperwork
• Discovery: San Francisco police need to fill out arrest-level paperwork just for talking to people
• Solution: Use AI and computer vision to reduce 2-3 hours of paperwork to 5-10 minutes

Finding Outsourced Jobs

Golden Rule: Any job outsourced to low-wage countries is an excellent AI startup opportunity right now.

Case: Lilac Labs - AI Drive-Thru Assistant

• Discovery: Many US drive-thru order takers actually work in BPO centers on the other side of the world
• Solution: AI completely automates drive-thru order processing

Method 4: Finding Products That Need Consultants

Case: Automat - RPA Revolution

• Market Pain Point: RPA products like UiPath are powerful but require expensive certified consultants
• Solution: Build truly usable RPA products without consultants
• Technical Foundation: Leverage latest AI advances, enabling LLMs to browse web and use desktop applications

Practical Tips: Rapid Opportunity Discovery

1. Indeed.com Search Method

• Search keywords: "remote analyst," "remote clerk"
• Look for jobs most people have never heard of
• These are typically targets AI can easily automate

2. Shadow Following Method

Find friends with boring jobs and follow them for a day.

Case: Sweet Spot

• Original Idea: Payment solutions for taco trucks
• Turning Point: Following friend's job—refreshing government websites all day looking for bidding opportunities
• New Direction: AI-powered government contracting and procurement platform
• Features: Find opportunities, generate bids, optimize pricing
• Growth: Growing incredibly fast

3. Technology Frontier Observation Method

Paul Graham's Advice: Live at the edge and notice what's missing.

Case: PreDB

• Opportunity Discovery: Friend needed real-time sync between PostgreSQL and Pinecone
• Deep Research: Discovered PG Vector extension could accomplish most Vector database functionality within PostgreSQL
• Achievement: Replaced both independent Vector databases and Elasticsearch

Avoiding Common Pitfalls

1. Over-Competition Phobia

Don't abandon good ideas because you see two competitors raise funding on TechCrunch.

Case: Gigl

• Competitive Environment: Customer support AI field looked very crowded
• Success Key: Technical strength, actually delivering what others cannot achieve
• Result: Secured contracts with major enterprises like Zepto

2. Investor Misunderstanding Syndrome

When you're out talking to real customers and getting orders, don't let investors' "lack of understanding" shake your confidence.

YC Partners' Advice:

3. Ideas Too Small Problem

Use the "squint test": If you 10x your idea, how would the world change?

Special Opportunities in the AI Era

Why Now Is the Best Time?

1. Dramatically Increased Value: Previously making software for small dental practices or auto loan companies might have limited value, but now using AI to replace human work, each position is worth $60-80K/year

2. Rapid Technology Development: New possible ideas emerge every few months

3. Sustained Team Morale: Working in the most exciting technology field keeps teams going longer

4. Higher Success Probability: Your lucky break might be just around the corner

Success Case Timeline

It's important to note that many of YC's most successful companies took about a year to find the right idea. This is completely normal.

The old notion that "either succeed immediately or seed extensions mean failure" is outdated. In the rapidly evolving AI era, persistence and exploration often lead to unexpected breakthroughs.

Action Guide: Starting Today

Immediately Actionable Steps:

1. Deep Introspection: List all your work experiences, internships, projects, looking for unique professional intersections
2. Activate Family Networks: Ask all friends and family about their work content, looking for pain points
3. Indeed Search: Search keywords like "remote analyst" to understand automatable jobs
4. Technology Frontier: Experience the latest AI tools and models, thinking about missing links
5. Field Research: Choose an industry of interest and find ways to understand it deeply

Mindset Adjustments:

• Reject Lazy Ideas: If you can build it over a weekend, it's probably not good enough
• Embrace Ambition: Make products slightly hard to build the first version
• Maintain Patience: Good ideas need time to discover and validate
• Trust Intuition: When customers are willing to pay, don't be swayed by external doubts

Conclusion: Get Out and Embrace the Future

YC Partners' Ultimate Advice:

In this golden age of AI entrepreneurship, opportunities belong to those who dare to explore, think deeply, and take continuous action. Whether you choose which path—mining internal expertise or exploring the external world—the key is starting to act.

Your next billion-dollar idea might be in the next conversation, the next observation, the next "what if we use AI to solve this problem?" thought.

---

This article is compiled from Y Combinator's latest sharing, providing practical guidance for AI entrepreneurs. For more entrepreneurship insights and data analysis, please follow our platform.

Related Resources:

• YC Startup Ideas Guide
• Paul Graham: How to Get Startup Ideas
• AI Startup Trends Analysis Report