Archive
2026
- The new software
LLM-written code should not simply mimic human-written code
- Untangling the meaning of "agent"
When is an agent not an agent?
- LLMs and programming language diversity
LLMs don't seem to be leading to a software monoculture after all
2023
- AI aligment by analogy
A brief sketch of AI alignment as an engineering problem
- Network states and eudaimonia
Only states whose citizens are flourishing deserve to survive
- AI in the cozyweb
From the all-seeing eye of Google, to the AI in every chat room
- Models are cultures
Traditional software is culturally-unaware, but deep learning models can grasp more of human meaning
- The economics of due diligence
Why did so many people get scammed in crypto?
- LLMs are hustlers
Computers that can't talk to each other can't trade with each other, but LLMs are good at talking