From a four-hour investor meeting with DeepSeek founder Liang Wenfeng， first reported by the WeChat publication elsewhere：
- Product is just a step on the way to AGI. When you build lower-level tech from a higher technical position， it's overwhelming force. Product is a byproduct.
- AI today doesn't lack taste or intuition. It lacks the ability to keep learning.
- DeepSeek's long-term vision is AGI. If the path is a staircase， last year's step was CoT. This year's step is Agent. After Agent， the problem is continual learning.
- Once models can keep learning， we may reach a gradual singularity： AI doing everything humans can do， including building better AI. AI accelerating AI research. Only after that comes embodiment.
- The first goal of our models is not that everyone finds them useful. It's that we find them useful - for building the next model. That's the fastest way to AGI.
- Selling APIs just isn't that attractive to me. A few people to maintain it， no customer support， no sales. Users come on their own.
- AI is big enough that it may end up taking ten percent of human GDP. If we try to keep that for ourselves， history will leave us behind. That's just how history works.
- I'm not worried about people deploying our models to compete with us. Startups are too small to do it. Big companies can't organize for it. Open source is the sweet spot for a company of our size.
- We want to rewrite the China-US AI story： close the gap to six months， three months， with a fraction of the compute. The gap is in resources， not talent. The talent is the same people.