China Just Picked Up an OpenAI Researcher — and a New Submarine

Two stories dropped this week that don't seem connected. They are.

On the AI front: Yao Shunyu, a former OpenAI researcher, is now the chief AI scientist at Tencent — one of China's largest tech companies. He made his ambitions public at a Tencent event in Beijing co-organized with local government authorities, where a senior Beijing official gave opening remarks. This wasn't a corporate conference. This was a state-adjacent rollout.

Yao's stated goal: build AGI — artificial general intelligence, meaning AI at or above human-level capability — inside China.

"My personal goal is that in China we should establish a long-term AGI organization," Yao said, according to CNBC's translation from Mandarin.

The AGI Ambition Shift Is Real

For years, Chinese companies acknowledged they were behind on AGI and focused on practical AI applications — factories, consumer electronics, productivity tools. Baidu CEO Robin Li publicly predicted AGI wouldn't arrive until at least 2034.

That posture is changing. As Chinese firms recruit Silicon Valley talent, they're importing the AGI vision along with it.

Yao is not the only case. Per CNBC, other Chinese tech companies have also been pulling talent from Silicon Valley in recent months. He said at the Beijing event that the untapped potential in AI is in the "trillions of dollars" and that ChatGPT and Claude will not be the only dominant super-apps.

His roadmap: foundational knowledge, frontier research, and real products — with a focus on smaller, more reliable AI models tuned for consistent performance on basic tasks.

While China Accelerates, Anthropic Wants a Pause

The same week Yao announced China's AGI ambitions, Anthropic — the San Francisco-based AI safety company — warned that frontier AI models are approaching the point where they can improve themselves without human oversight. Anthropic called for an industry slowdown or pause in new model development.

Anthropic has also separately urged Washington to maintain the U.S. lead over Chinese AI models.

The same company urging the U.S. to stay ahead is also urging the industry to slow down. China, meanwhile, is not slowing down. They're hiring.

The timing creates a real strategic problem that mainstream coverage is largely glossing over, even if the safety concerns themselves are legitimate.

The Submarine Nobody Can Identify

Separate from the AI story — and equally significant — fresh satellite imagery has revealed a large new submarine at Shanghai's Jiangnan Shipyard.

According to Interesting Engineering, via ZeroHedge, the vessel is approximately 120 meters long. Its defining feature: no sail. Conventional submarines use the sail structure to house periscopes, communication masts, and snorkel systems. Remove it, and you get less drag, better hydrodynamic efficiency, and a quieter acoustic signature.

The result: harder to find.

The hull also features X-shaped stern control surfaces and what analysts believe may be a shrouded propulsion system — possibly a pumpjet propulsor, which reduces noise at higher speeds compared to standard propellers.

The propulsion source remains unconfirmed. Given the size of the vessel, defense analysts believe a nuclear reactor is the most likely option. China's development of air-independent propulsion systems — a non-nuclear, quieter alternative — is also a possibility that analysts haven't ruled out.

Jiangnan Shipyard tested a smaller experimental submarine with a reduced sail design roughly eight years ago. This new vessel appears to be a significantly scaled-up iteration of that concept.

China Has Launched 15 to 20 Submarines in Five Years

According to Interesting Engineering, China has launched roughly 15 to 20 submarines over the past five years, several of them belonging to entirely new classes. The U.S. and its allies, meanwhile, are struggling to increase submarine production output.

The U.S. Virginia-class submarine program has faced well-documented delays and cost overruns. The AUKUS agreement to supply Australia with nuclear submarines won't deliver boats for well over a decade. The production math is not in the West's favor right now.

The Full Picture

CNBC's AI piece focuses on the talent drain story accurately, but frames it primarily as a business and technology narrative. The national security dimension — a state-adjacent Beijing event, government officials on stage, coordinated recruitment of Western AI researchers — gets soft-pedaled.

ZeroHedge's submarine piece aggregates the Interesting Engineering report accurately but, as usual, leads with maximum alarm framing. The facts are alarming enough on their own without the theatrics.

These two stories are the same story. China is running a coordinated, state-backed effort to close the gap with the United States across both the intelligence and military domains simultaneously. AI and submarines are both pieces of the same strategic competition.

What It Means

If you're a taxpayer, this is where your money should be going — and largely isn't. The U.S. defense industrial base is producing submarines too slowly. The AI industry's best safety-focused researchers are publicly debating slowdowns while Chinese companies actively poach their colleagues.

Washington can argue about budget reconciliation and pronoun policies. Beijing is building submarines nobody can identify and hiring the people who built ChatGPT.

The scoreboard updates whether anyone's watching or not.