Matt White spent the last week of April 2026 visiting the labs behind most of China's frontier AI work — DeepSeek, Moonshot, Z.ai, Alibaba's Qwen team, MiniMax, ByteDance Seed, Ant Group, 01.AI, Xiaomi — along with several robotics startups and Tsinghua's AIR Lab. Five of them shipped open-weights releases that same week, including DeepSeek's 1.6-trillion-parameter V4 the day White landed.

He argues that U.S. export controls have not slowed Chinese AI so much as redirected it. The compute constraint has pushed the labs toward more efficient architectures and homegrown chips, with Z.ai now training entirely on Huawei Ascend and DeepSeek moving in the same direction. The labs themselves are small, young, and culturally committed to open source in a way that is harder to find at American frontier labs. Chinese graduates who would have defaulted to a U.S. lab two years ago are going home, and increasingly are not leaving in the first place.

While White was on the ground, the House Foreign Affairs Committee advanced the Deterring American AI Model Theft Act, China's NDRC told several leading labs to reject U.S. investment without government approval, and Beijing unwound Meta's completed acquisition of Manus. What White saw inside the labs that week complicates the story both governments are telling about each other — and points to a third path that neither has yet taken seriously.

Join us for a full-day, in-person build session to get hands-on with OpenAI’s next-generation stack, including Workspace Agents, Codex, Realtime API, and the Agents SDK. Whether you are a seasoned technical builder or a first-time hacker, this event is designed for you to ship real projects.

What you'll get:

⚠️ Claim a BOGO seat at OpenAI Business NOW— don't wait until the day of.

Anthony Garcia, one of the early builders of The AI Collective, posted a reflection this week on the three-year mark. The original dozen people who met in a San Francisco living room now sit inside a community of 250,000+ members, 200+ active chapters across 40+ countries, with 500+ startups showcased through demo nights, pitch events, hackathons, panels, and private dinners.

His takeaway was unflashy. The number that matters is not the headcount. It is the founders building futures that do not yet exist, the creators imagining new worlds, and the small group who decided three years ago that great things often begin with conversation, community, and bold ways of thinking. That framing is the one to keep when the next 250,000 join.

The Boston chapter hosted Genspark for the Boston leg of the Genspark Claw World Tour, a five-city global activation pairing AIC members with the Genspark team for a live demo and hands-on build session. Doors opened at 6 PM with networking, followed by an intro to Genspark Claw, a live demo, an interactive workshop where attendees worked through Claw on their own laptops, and a show-and-tell where teams shared what they had shipped during the session.

Genspark sent Nellie Taratorin and Mariana Klober to lead the program, joined by AIC Boston hosts Janna, Kanal Patel, Ananya Shah, and Jason Morris. The pitch from Genspark is positioning Claw as the enterprise-ready option for teams that want comparable capability with tighter security guarantees. Builders who want to see how that holds up in practice should keep an eye on the remaining tour stops.

Evaluating enterprise AI on standard English benchmarks creates a dangerous illusion of global readiness. Because flagship models heavily over-index on English training data , standard monitoring systems frequently miss cross-lingual contextual drift, cultural edge cases, and silent failures in non-English outputs. Relying on English-first tools means flying blind post-deployment. True global reliability requires a continuous, native-language benchmarking infrastructure that treats every language as core engineering infrastructure.

Our partners and sponsors get exclusive placements across the newsletter and access to AIC's in-person network — demo nights, dinners, hackathons, and forums across 180+ chapters.

For all inquiries, send us a note at [email protected].

The AI Collective is built by volunteers across 180+ chapters in 40 countries.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.