Now

This is a now page. Last updated: January 2026.

---

Things I’m excited about

1. Claude Code and its implications
2. Privacy tech and its implications

People I’m taking seriously

1. David Deutsch
2. Daniel Everett
3. Li Lu
4. Warren Buffett
5. Ludwig Lachmann
6. Paul Graham
7. Murray Rothbard
8. Eric Voskuil
9. Scott Aaronson
10. Naval Ravikant
11. William Zinsser
12. Michael Burry

Essays I’m thinking about

1. You shouldn’t hold more than 15 names in your portfolio. You shouldn’t be working more than 15 projects at the same time. I also think you can’t take more than 15 people seriously at the same time.
2. Write what you wouldn’t dare say
3. Naruto and LLMs
4. Elon says everything will be almost free with AI. But is that true? What’d be the implication?
5. Not all corporate networks are coercive (imperial) networks. And it’s not nation-states per se but coercive networks with soft power (or hard) we should be paying attention to. You have to understand geopolitics, history, biophysics, tech, economics, and epistemology—and how they relate to each other.
6. Maybe Warren Buffett was corporate network investor.
7. Maybe I’m using bullet points too much
8. Daniel Everett used Peircean semiotics—icon, index, and symbols. I think there are few more steps in between each: particularly, we can de-index something without giving the thing corresponding symbol. This is how privacy (i.e., unidentifiable) is achieved. And I think this is topologically similar to how fungibility defines the multiverse before entanglement—put differently, in my mind the process of decoherence resembles the transition from de-index to symbol.
9. I think L1-L2 architecture is good analogy and should be pursued. I was taking Elie Ayache’s argument too seriously—it’s true that L2s could become L1s, but L1s and L2s both exist. By definition, L2s (e.g., derivative market) are dependent on L1s (e.g., the underlying market). The free market force will coordinate the size of L2 if it gets too big, similar to how the triad will be aligned within free market society. It can stay too big only because it’s dependent on nation state networks. Li Lu’s emphasis of the Iron Law of Civilization 3.0 should be interpreted in this context—i.e., you must know where the L1s are, and acknowledge L2s’ dependency on them. Remembering this dependency is the difference between localization and fragmentation—we should strive for the former, but never the latter. The tricky part, which is related with #5 above, is that coercive network tends to intervene such local development. Li Lu’s emphasis on real wealth is understandable in this context—if the L2s you belong see themselves as L1s, they will take coercive measure to lock-in the participants (e.g., via taxation).
10. I think we are getting cleaner view of what exactly constitutes moat given recent development in AI. Maybe the moat is less technical and more legal. Inspired by this.
11. I feel like CPU and GPU architecture (not just LLMs), as well as its history and evolution (and also what’s powering computers and mind), can teach us a lot about how mind works, since mind works via computation. If LLMs can inform us about the software, then CPU and GPU can inform us about the hardware. (<> 8-2b2d Value shifts to adjacent layers)
12. What makes a good summary?

---

Back to About this project