ChinaTalk is a Substack publication and podcast run by Jordan Schneider covering Chinese technology, US-China relations, and AI policy. As of 2025, the publication reported around 50,000 Substack subscribers and was building out a "ChinaTalk Institute" think-tank arm with full-time staff. Its archive runs to hundreds of pieces, including the "DeepSeek's Edge" essay (co-authored with Lily Ottinger and Kevin Xu) and a sequence of analyses on US export controls, the CHIPS Act, and the BIS AI Diffusion rule. Lily Ottinger, a ChinaTalk editor, authored "The Cold War History of Export Controls" and other long-form posts that became reference material in the US semiconductor-policy debate.
The Manifund regrant from June 2024 supported deeper coverage of Chinese AI policy and the DeepSeek lab specifically. Earlier Manifund regrantors Joel Becker and Evan Hubinger had together routed approximately $37,000 to ChinaTalk in 2023 and 2024; the 2024 regrant funding the DeepSeek-focused coverage came from the same regranting pool. ChinaTalk published technical and policy coverage of DeepSeek's V2 and V3 lines through 2024, before DeepSeek-R1's January 2025 release brought Chinese open-weights frontier labs into mainstream US discussion.
In the open-source AI stack this work sits at the governance layer. It functions as primary-source reporting on the Chinese end of the global open-weights flow: which labs (DeepSeek, Qwen, Moonshot, Z.ai) are releasing what, how Chinese chip supply chains adapt to US controls, and how Beijing's policy stance toward open-weights frontier models is moving. The Manifund retrospective cited ChinaTalk's pre-R1 coverage as ahead of where the policy debate later landed.
Recipient
ChinaTalk (Jordan Schneider)
Funder
Manifund · foundation · US
Operates an AI safety regranting program that gives expert regrantors $100K+ budgets to make fast, low-friction grants to early-stage technical and policy projects.
Primary source
Additional sources
More from Manifund
- Frontier AI benchmark pilot 2024-09-01 · $200K
Regrant from Leopold Aschenbrenner supporting Epoch AI's pilot of a new frontier AI benchmark, building on their earlier compute-trends and capability-tracking work.
- Scoping Developmental Interpretability (Timaeus first funding) 2024-03-01 · Undisclosed
First funding for the Timaeus team's developmental interpretability research program. Manifund regrant accelerated their research by months and seeded what is now an established alignment lab.