Core dump

By default, anthropic hides these "inner thought" aka text chunks. This is the reason why sometimes you see the anthropic flower loading in the middle of a response. Claude is basically deliberating at arbitrary points of a response for a better output - imo one of the most underrated unhobblings.

Comparisons are sometimes made between LLMs and calculators, to make the point that AI bans are as futile and philistine as calls to return to the abacus. But the work that we bypass when using a calculator is less important than what we bypass when using an AI language generator for writing. To be a human self, a human agent, is to be a linguistic... See more

For this reason, I suspect that many of my friends who write and publish rapidly are shortchanging themselves. They generate texts filled with hidden doors and move on before they’ve opened them.

When I try to think about thinking, for instance retracing where an idea of mine came from, the limitations of English force me to say that "I" "produced" an "idea." But none of these things are stable entities, and this grammatical relationship among them is misleading. The "idea" isn't a finished product wi

... See more

When I write, I get to observe the transition from this fluid mode of thinking to the rigid. As I type, I’m often in a fluid mode—writing at the speed of thought. I feel confident about what I’m saying. But as soon as I stop, the thoughts solidify, rigid on the page, and, as I read what I’ve written, I see cracks spreading through my ideas. What se... See more