Y Combinator builds AI versions of its partners to talk through startup ideas

Wafer says Y Combinator built its Office Hour Simulator to help more people work through startup ideas with AI versions of YC partners. According to Wafer, YC compared GLM-5.2 on a dedicated Wafer endpoint with GPT-4.1 mini on OpenAI and Gemma 4 31B on Cerebras, evaluating answer quality, latency and conversation duration. Wafer reports 31% lower average language-model latency than the OpenAI configuration and 44% lower than the Cerebras configuration. It also says users spent 2.5 minutes longer talking to YC’s AI partners when using Wafer.