Chang Xu
Partner at Basis Set. I invest in and build AI agents.
About
I invest in and build AI agents at Basis Set. I studied Applied Math and Computer Science at Harvard, and took classes in ML before it was popular. I like to stay abreast of the research and learn the latest technologies and products.
Before venture, I was a founder and operator. I co-founded Onion Academy, an edtech company in China that has reached 130M users and raised $90m to date. I was also the first product manager at Minerva Project, a venture-backed, tech-enabled university of the 21st century. I joined Basis Set during fund I, and now we are investing $850M AUM across four funds. I started my career at BCG, where I advised clients across technology, retail, healthcare, and private equity. I have also spent time at Microsoft and Jane Street Capital.
A natural builder of mental models, I grew up loving math competitions. I built intuition with brute-force methods to uncover patterns, ultimately crafting the most elegant solution. This approach mirrors the startup journey: start with what doesn't scale, iterate, automate, and dominate. In my downtime, I enjoy strategy games and once solved a 266-piece all-white puzzle with my husband during COVID lockdown.
On the Radar
I run agents that continuously track AI research. On the Radar is their output: a running map of leading AI researchers by field —
world models,
robotics & physical AI, and
deepfake detection.
Elsewhere
Writing, talks & gatherings
- The AI lottery machine (Note, Jul 7, 2026)
- Is there an AI bubble? (Talk, Jun 23, 2026 · TechCrunch article)
- Takeaways from the first ACM Conference on AI and Agentic Systems (Event, May 29, 2026)
- Investing in the open-source agent stack (Essay, Apr 23, 2026)
- The future of agents is open source (Essay, Apr 20, 2026)
- Basis Set AI infrastructure founders dinner (Event, Mar 3, 2026)
- AI platform for cybersecurity (DeFUD) (Podcast, Feb 10, 2026)
- Infrastructure for the agentic era (Essay, Jan 25, 2026)
- Building with agents, not code (Essay, Jan 15, 2026)
- AI-first security design reviews (DeFUD) (Podcast, Jan 12, 2026)
- NeurIPS 2025 women's lunch (Event, Dec 11, 2025)
- NeurIPS 2025 podcast (Podcast, Dec 10, 2025)
- NeurIPS demo night (Event, Dec 10, 2025)
- AI-first security operations (DeFUD) (Podcast, Dec 10, 2025)
- On quantization and token economics for using frontier models (Note, Nov 5, 2025)
- On the economics of Claude (Note, Oct 15, 2025)
- It's time for a security foundation model (Essay, Aug 14, 2025)
- Black Hat security brunch (Event, Aug 8, 2025)
- How to build a financial model for your data room (Essay, Jul 8, 2025)
- Bad ideas are good again (Essay, Aug 15, 2024)
- All that glitters like stars is not gold (Essay, Apr 10, 2024)
- Dev Founders technical founders dinner (Event, Sep 22, 2023)
- How to commercialize open source (Essay, Sep 17, 2020)
- The future of college — TEDx (Talk, Dec 12, 2018)
- Cameras and cats (Talk, Sep 25, 2018)