The Y Combinator Request for Startups, from Spring 2026: 10 ideas YC’s partners want founders to build. Each links to a problem page with more detail, related problems and people working on it.
Language models struggle with geometry and physical structure. The ask is large spatial reasoning models that treat 3D structure as a first-class primitive.
False Claims Act cases recover money from companies that defraud the government, but building them takes months of manual work. YC wants software that turns an insider tip into a complaint-ready case.
Training language models still means broken SDKs, flaky GPU instances and wrangling terabytes of data. The ask is products that make model training easy.
AI coding tools are great once it's clear what to build. YC wants an AI-native system, a Cursor for product management, that helps teams figure out what to build in the first place.
The biggest funds have been slow to adopt AI. YC thinks the hedge funds of the future will use AI to invent entirely new strategies, not bolt it onto old ones.
Agencies have always been hard to scale: low margins, slow manual work, growth only by hiring. YC wants AI-native agencies that sell the finished work and run with software margins.
The GENIUS and CLARITY Acts put stablecoins between DeFi and TradFi: compliant but crypto-native. YC wants the financial services layer that is still missing on top of them.
AI is helping people fill in government forms faster, but many agencies still print and process them by hand. YC wants AI tools that help government handle the coming surge.
US buyers of rolled aluminum or steel tube often wait 8 to 30 weeks, because mills run on decades-old systems. YC wants modern, software-defined American mills.
AI can't do physical work yet, but it can see, reason and guide the person who does. YC wants real-time AI coaching that makes field, factory and healthcare workers effective right away.