Reactions from ranked influencers
3 posts
YC just dropped their Fall 2026 Requests for Startups. My takeaways: 1. Pick a business where operating it generates data nobody else can get. The software is the wedge. The data is the company. 2. Almost none of these ideas are apps. They're plumbing. A cloud for small software, self-maintaining APIs, escrow between machines, compliance rails. Kinda interesting. 3. 3 of 13 are about proving something is real. 25% of the list is dedicated to the fact that it's really hard to decipher whats true/real anymore. 4. The best line is in the physical world section. Build the software that runs construction crews and you end up owning data on how the real world actually works, 5. The wedge and the ambition don't have to match. The Primer starts as a reading app a parent buys and ends as something that raises a kid for a decade. Small entry point, enormous end game. 12 more AI native startup ideas based on where YC says the world is going: 1.Moving companies off their old software. Every business needs to get agent-native and dreads the project. Charge $50k, let agents do 70 percent of the work, run 80 percent margins on something every company needs once. 2. An AI tutor for adults getting replaced at work. The paralegal watching her job disappear pays $300 a month because this is her career. Get 100 people paying, that's $360k a year and your only cost is tokens. 3. Insurance for when agents mess up. An agent books the wrong flight, a robot dents a truck. Nobody's underwriting that yet. You broker it, take 15 percent of premium, and own the category before the big carriers wake up. 4. Proof that the job got done. Photos, sensors, timestamps that settle disputes between a contractor and a homeowner. Home services is $600 billion and disputes eat 3 to 5 percent of it. Charge $200 a month per crew, land 500 crews, that's $1.2M a year. 5. A card your agent can use that isn't yours. Spins up per task, dies when it's done. You make 1 to 2 percent on everything that flows through. Customers pushing $10 million a month means $150k monthly while you sleep. 6. Contractor licensing in one state. Texas has 60,000 licensed contractors in renewal hell. Charge $2k a month, sign 500 of them, that's a $12M business without ever leaving Texas. Then do Florida. 7. Escrow for agents paying agents. Hold the money, release it when the work checks out, keep 1 percent. Sounds tiny until agents are running millions of transactions a day and you're the toll booth. Thanks to @ideabrowser for this idea (more ideas there). 8. A record of why your agent did what it did. Hospitals and banks refuse to deploy agents without this. Compliance software sells for $50k to $200k a year per company because the fine costs ten times more. 9. Practice software for dentists, rebuilt for agents. 200,000 practices all running software from 2011. Charge $150 a month, get 10 percent, that's $36M a year in a category that isn't popular with VC funded startups. 10. Software for the daughter managing her mom's care. 53 million Americans do this unpaid, coordinating meds in group texts. Charge $30 a month. This market grows every day for 20 years and right positioning/marketing/brand can get your far here. 11. The agent that reads every contract your company already signed. Most businesses have hundreds sitting in a folder nobody's opened since signing. Find the auto-renewals, the missed rebates, the terms you're paying for and not using. Charge a percentage of what you save them and it sells itself. 12. Grading the agents you already run. Once a company has ten agents working, they need to know which ones are actually good. Sell the scoreboard. Nobody buys this until they're deep in, which means the market shows up in about a year. Happy building, friends.
AI is moving into the physical world. We're excited about a new wave of startups rebuilding the systems that power the real world, from education and healthcare to defense, finance, infrastructure, and work itself. http://ycombinator.com/rfs
Y Combinator's Fall 2026 request for startups:
it's funny because i am kind of doing all of the top 4 here
Y Combinator's Fall 2026 request for startups: