The state of LLMs is messy: Some AI features (like vision) lag others (like tool use) while others have blind spots (imagegen and clocks). And the expensive "heavy thinking" models are now very far ahead of all the other AIs that most people use. None of this is well-documented.
@emollick
-
Debating Core Analogies: Understanding AI’s Impact and Nature
By
–
I guess I would be remiss for not including other analogies that get debated here: the eschaton or the home computer? The atom bomb or crypto? A child or a plagiarism machine?
-
AI Analogy Problems: Disagreements About LLM Nature and Impact
By
–
Many of the disagreements over AI are analogy problems: is an LLM a brain or parrot? Will its effects on productivity be electrification or internet? Is the build-out of AI capability like the cloud or the 1880s railroad boom-and-bust? No one analogy fits, so we fight over them.
-
The Switch That Controlled Early ARPANET Internet
By
–
This switch turned off 1/3 of the Internet. Or at least it did in the earliest days of ARPANET in 1970, where it controlled the key BBN node in Boston. For better or worse, it no longer works (I tried flipping it when I visited the company).
-
Sora 2 generates elaborate music videos from simple text prompts
By
–
All of the dialogue and staging was from Sora 2. For example the music video prompt was just “1980s over-the-top music video about an otter sitting in an airplane seat using a laptop”
-
Radiologists and AI displacement: Evidence from O*NET studies
By
–
This is not a criticism of @binarybits who covers this stuff in a sophisticated way. But I think radiologists were always a naive target for this sort of speculation. We know a lot more from the O*NET studies about potential impacts and are starting to see them in the real world
-
AI Automation: Policy Needed Despite Hinton’s Optimism
By
–
I think that that it is good to push back against naive assumptions about automation, but I worry Hinton’s bad prediction provides too much comfort. AI scientists are bad predictors of downstream effects, but that doesn’t mean that we don’t need policy to mitigate AI job impacts.
-
Sora 2: Simplified Video Creation Interface for Viral Short-Form Content
By
–
Not a good interface for “real” video creation compared to Sora 1 (only one video at a time, limited controls, hard to do selection and variation). But it feels exactly like a short video app & thus accessible for viral use.
-
Sora 2 Launch Strategy: Gated Access Drives User Growth Virality
By
–
The labs learned from the Studio Ghibli thing that images & video could produce viral moments that turn into user gain. The Sora 2 launch is the ultimate implementation of this: gated invites, an app that selects for virality, reasons to share with friends, provocative content
-
World Models Learned from Video: New Evidence and Implications
By
–
More evidence that world models can be learned from video alone (and a small amount of video at that): https://
arxiv.org/pdf/2509.24527