The Y Combinator Request for Startups, from Spring 2025: 12 ideas YC’s partners want founders to build. Each links to a problem page with more detail, related problems and people working on it.
Most software assumes a human user, so bots end up posing as people. YC wants services designed for AI agents as the customer.
Open source tends to produce companies that sell support around it. YC wants startups that help businesses adopt and run open-source AI.
Reasoning models shift compute from training to inference, and costs climb as apps make far more calls. YC wants a cheaper, better inference stack.
YC expects a wave of vertical AI agents, LLM software tuned to fully automate specific jobs, to rival the B2B SaaS wave of 2005 to 2020.
Nvidia's lead rests largely on hand-optimized CUDA code. YC wants AI that writes kernels and other low-level code so alternative chips can compete.
YC expects AI to push the cost of building software toward zero, with engineers managing teams of agents instead of writing code. It wants the tools for that.
As reasoning models let AI agents act on their own, YC wants the tools to build them: agent builders and building blocks that extend what agents can do.
Wealthy people employ accountants, lawyers, trainers, tutors and doctors. YC wants AI that brings that kind of personal staff to everyone.
AI agents can now use websites and desktop apps. YC argues this effectively turns every site and app into an API, and wants founders to build on it.
About 4 million people in the US and Europe work in compliance and audit, much of it manual reading and sampling. YC wants LLM tools that review everything, continuously.
AI progress needs more data centers, built faster and cheaper. YC wants startups that speed up the buildout, from power and cooling to procurement, project management and autonomous operations.
YC wants an AI app store and OS layer on phones and computers that keeps data under the user's control, holds one shared memory, vets apps and handles payments.