March 2026 Links
Some Interesting Things I Read (and Watched) This Month
-
My Thoughts on the Future of "AI"
— Here Nicholas Carlini, a researcher at Anthropic, argues that we should exercise humility when speculating on the future of AI. That is, recursive takeoff and a hard limit are equally likely, but improbable tails — somewhere in the middle is most likely. As someone whose opinion tends toward the hard-limit side (and it seems Carlini, or at least people in his circle, tend to as well), I appreciated the special attention he gives to arguing against this position.
“So over the next few years, I'd encourage you to keep an open mind and be willing to see the world as it is, and not as you want it to be. We're going to learn a lot, things are going to change a lot, and so we need to be willing to accept what comes to pass, and not reject something just because it's not what we expected.”
- Every minute you aren't running 69 agents, you are falling behind — This post from George Hotz presents a view on the utility of AI, the toxicity of doomer social media, and the likely effect of LLMs on our economy — a consolidation of rent seekers. He also argues that labor which actually creates value for people is the least likely to be disrupted. This is a view I largely endorse. If your job primarily creates friction for others, you are likely to be steamrolled by agents. If you create physical goods or provide services people genuinely want, your role is far more robust in the face of this disruption. Geohot may place more moral weight than I would on certain jobs that fall into this latter category (for example, I think both novel medical R&D and operating an online casino fall into the “robust” category, though one is clearly more righteous), but I broadly agree with the framework.
- LLMs can unmask pseudonymous users at scale with surprising accuracy — Pseudonymous social media users were able to be “unmasked” at rates and accuracy higher than previous techniques by using an LLM to correlate posts and online presence across platforms. As someone who has worked in the privacy space, this result is not particularly surprising to me, but it is still quite interesting.
- My Longest Ever Flight: Cathay Pacific from Boston to Hong Kong! — A YouTube video from Miles in Transit about his recent experience flying to Taiwan. I’m highlighting this video for two reasons. First, having just flown to Taiwan and back myself, I relate to the trip. Second, it’s a chance to highlight a great YouTuber in a niche space: transit. Miles’ videos about long Greyhound trips, seldom-used Amtrak stations, and generally pushing public transit to its extremes are excellent, and I encourage people to check out his channel.