No. 1
27/10/2026 · 8 pieces
8 picks from 146 pieces reviewed in 10 publications across 3 regions.
-
01 ★★★★★
How Chinese companies talk about recursive self-improvement
What do Chinese AI companies mean when they talk about AI improving itself? Zilan Qian and Cheryl Wu read about 1,400 company documents to find out. Read it to test the common story of a US-China race toward recursive self-improvement against what Chinese firms actually say. We picked it for original data on a question usually argued from assumptions. Qian and Wu write for Asterisk, a magazine on science and technology; we found no stake.
-
02 ★★★★★
MOFCOM's New Rules and the Future of Chinese AI
Will China's new export-control rules end the open release of its AI models? The ChinaTalk team, led by Quinn Ennis, spent three days on the commerce ministry's draft and explains what it would cover. Read it to understand a change that could reshape the open-model world, and the question still open at its centre. We picked it as a close reading of a primary text that most coverage only summarised. Ennis and colleagues write for ChinaTalk, a newsletter on China and technology; we found no stake.
-
03 ★★★★☆
OpenAI's ironic credibility problem
When OpenAI fired three safety researchers, how could anyone inside know the real reason? Dan Kagan-Kans does not try to settle who is right, and asks instead what the episode means for employees who may one day have to raise concerns. Read it for a sharp point about trust and verification inside AI labs. We picked it for turning a news story into a lasting question. Kagan-Kans writes for Understanding AI, a newsletter on AI and its effects; we found no stake.
-
04 ★★★★☆
A big-tent or small-tent AI safety movement?
Should the AI safety movement focus on extinction risk, or on the many smaller harms already in sight? Arvind Narayanan and Sayash Kapoor make the case for a broader movement while conceding much to the other side. Read it for a fair-minded argument that clarifies what the safety debate is really about. We picked it because it engages its opponents instead of caricaturing them. The authors are computer scientists known for scepticism about AI hype and existential risk, a position stated openly; we found no other stake.
-
05 ★★★★☆
The current balance of power in open models
Who leads in open AI models, the United States or China, and does it matter? Nathan Lambert sets out the evidence prepared for a briefing to members of Congress. Read it for a compact, numbers-led view of the open-model race and of what restricting Chinese models would cost. We picked it as a primary text from a policy briefing. Lambert builds open models for a living, a stake the piece discloses and readers should weigh.
-
06 ★★★★☆
Why humanoid robots won't catch up to human workers any time soon
Why do humanoid robots that dance and sprint still struggle to peel an orange? Kai Williams talked to many roboticists about the gap between viral demos and useful work. Read it to understand which problems are solved, which are not, and why the hard ones could take years. We picked it for deflating hype with reporting rather than opinion. Williams writes for Understanding AI; the piece carries a sponsorship from 80,000 Hours, a careers non-profit, and we found no stake in robotics.
-
07 ★★★★☆
What will happen to Indian IT?
What happens to the world's back office when AI can write code and answer calls? This Asterisk essay traces India's bet on exporting skilled labour, from IBM's exit in 1978 to an industry employing millions. Read it to understand how much of India's economy rests on an arbitrage that AI now undercuts. We picked it for a long view of a question usually asked only about Western jobs. The essay appears in Asterisk, a magazine on science and technology; the page we read does not name the author, and we found no stake.
-
08 ★★★★☆
The Lithography Lever
Where is the West's real leverage over China's AI chips? In this ChinaTalk interview, Jordan Schneider presses Nick Brown of the Center for Technology and Statecraft on lithography machines, compute estimates and the proposed MATCH Act. Read it for a detailed, numbers-first view of what export controls could still achieve and what they might provoke. We picked it for its depth on the one input that is hard to replace. Both argue for tighter export controls, a position stated openly; we found no other stake.
No pieces match this filter.