Manus' Final Interview Before the Acquisition: Oh, the Surreal Odyssey of 2025…

In late 2025, on the podcast Mingdu, host Xiaojun sat down with Ji Yichao (PiK), co-founder and chief scientist of Manus AI. Dubbed the ‘Year of the Agent,’ 2025 saw Manus rise to prominence with its general-purpose AI agent – and the person in its promo videos who was called ‘too normal’ was none other than PiK himself. The conversation unspools his entrepreneurial journey across the mobile and AI eras, mixing nostalgia, candor, and self‑deprecation.

PiK’s background is unusual. He moved to the US at four, returned to China in second grade; his father is a physics professor at Peking University, his mother a serial entrepreneur from Zhongguancun. In 2009, when Apple’s App Store launched, the high‑schooler saw an opening. He single‑handedly built a paid third‑party iOS browser and earned over $300,000 selling copies. That early windfall gave him both cash and conviction that ‘innovation pays.’ Soon, ZhenFund’s Xu Xiaoping offered him a term sheet with a promise: do whatever you want. Ji dropped out of university, assembled a small team, and started his first real company. But they quickly realized the pay‑per‑copy model was dying, so they pivoted to the fledgling field of natural language processing.

From 2011 onward, Ji’s team trekked a hard road. Sensing that traditional search would fail on wearables and voice interfaces, they set out to build a system that could automatically read the web, extract entities and relations, and continuously update a knowledge graph – no predefined schemas. They trained models from scratch, riding each algorithmic wave: Word2Vec, LSTMs with attention, and then contextual encodings. Each breakthrough forced them to discard years of work. By 2018 they had tackled long‑context (16K tokens), but everything changed in 2019 when they got early access to GPT‑3. Ji recalls feeling ‘the sky was falling’ – their custom end‑to‑end system was matched by a casually prompted general model. It dawned on him that fine‑grained task specialization would soon be subsumed by general intelligence.

Ji doesn’t just rue the pace of change. He acknowledges that his early success came from catching a ‘wild west’ moment of platform shift, something the AI era lacks. He also questions whether dropping out was wise, conceding that luck and opportunity cost drove the choice. He levels a sharp critique at the industry: many founders are too ‘artistic’ – obsessive, extreme, yet without Steve Jobs’ destiny, only his disease. He even half‑jokes that he himself gets called ‘too normal’ in the company’s promos. This self‑awareness gives his story extra weight.

By the end, Ji says he has ‘aged into a lamp’ and that his intellectual prime was spent on that knowledge‑graph project. Perhaps those humbling years are what keep him grounded amid the Manus hype. The trait he values most – being ‘normal,’ sound in body and mind, free of extreme thoughts or habits – seems almost rare in today’s AI frenzy. This sprawling interview is both one person’s autobiography and a group portrait of a generation of builders wrestling with a revolution that outruns their best efforts.