The more powerful our language models become, the more they resemble learned ignorance. The real breakthrough is domain mastery, not universal knowledge.
Most AI demos work perfectly. Most AI systems fail at scale. The problem isn't the model — it's everything around it.
Get new essays by email.