Attention — Oct 28, 2026
What caught my attention on Oct 28 (3 items).
NCM: The End of LLMs? Not So Fast (Local Tests)
Prompt Engineering (Article:10:28) - Published 2026-10-27 18:00 - Discovered 2026-10-28 10:28
The video examines Stanford’s Next-Concept Modeling (NCM), a language model that generates text by first producing a coarse sketch of an entire answer and then refining it in a small number of passes, rather than autoregressively emitting one token at a time. NCM compresses text into hierarchical learned codes, or “concepts,” and uses a transformer to predict one level of codes per pass. The paper…
The Truth About Data Centers w/ Paris Marx | Better Offline
Better Offline (Article:10:53) - Published 2026-10-27 18:00 - Discovered 2026-10-28 10:53
In this episode of Better Offline, host Ed Zitron speaks with Paris Marx, author of Hyperscale and host of Tech Won’t Save Us, about the rapid expansion of data centers driven by the AI boom and the social, environmental, and political consequences of that buildout. Marx explains that data centers are not new, but the generative AI surge has dramatically increased their scale, energy demand, and v…
The Chinese AI crash is not the end, but it is cheering
Pivot to AI (Article:17:08) - Published 2026-10-27 18:00 - Discovered 2026-10-28 17:08
David Gerard argues that reports of a bursting Chinese AI bubble are not a sign that AI is finished, but a welcome correction of speculative excess. He notes that while much attention focuses on US AI and venture capital bubbles, China has had its own stock-market bubble, especially around AI and semiconductor companies and humanoid robot firms such as Unitree. These companies have been hyped as t…