Y Combinator spent fifteen years funding software and almost no services. Its latest batch has nine AI-native service companies — an AI law firm, an AI accountant, an AI insurance broker — and a Startup School session teaching founders how to build one. The thesis is clean: don't sell software to accountants, do the accounting. I sat down with two founders in Singapore who arrived there on their own, eighteen months earlier, and who have already hit three walls the thesis doesn't mention.
Chris Liu was a product manager at Weibo and Alipay, then a three-time founder. He landed in Australia in 2024, went undercover in a furniture store, and spent a few days as a painter's assistant. On why Australia is the market everyone skips, why you can't sell AI tools to a country of small businesses, and why closing deals there is easier than in China.
Aria Wang went from ADM — one of the world's four largest agricultural traders — to INSEAD to building AI agents for food and beverage companies. On why tech and food can't understand each other, how AI-simulated consumers hit 90% correlation with human panels, and what she learned winning a Google DeepMind hackathon with no technical background.
Mao Hua spent eight years building Tencent's AI systems — from IoT to voice to computer vision. Now he's in the US, turning a finance agent into a full-service accounting firm. His take on why the cloud is about to get radically simpler, and why your MacBook will run your company's AI within two years.
Leo Jiang went from BT and Vodafone to building Huawei's Spark program — 7,000+ startup applications across five continents. Then he left to build an AI sales agent. On why outbound sales is the perfect AI use case, why SaaS is ending, and why the future is agent-to-agent commerce.
Hu Chenchen went from Tsinghua professor to Xilinx Asia CTO to NIO's chief expert to AI startup founder. His thesis: the last decade belonged to the cloud, but the next decade belongs to the edge. A medical AI device that costs 1/20th the power and 1/60th the volume just proved him right.