It Depends.

The title represents a most honest answer to any question.

Let's examine the correct response playbook. In order:

1. Why are you asking this question? Reflect on the question underneath.
2. What do you think the response might be to the question? Estimate a bit.
3. Where might we want to direct our attention next, to resolve dependency??

Repeating this process takes you to the root. It's not quite therapy, but it's not quite bullshit either.

Children stump adults on the meaning of life very quickly if they keep asking "but why?" like a broken record. It shows that the answer to any question is, at core, it depends.

Will the child understand my response, enough to stop being a broken record?

The adult version comes from George Mack, who will go down in history as the man who made HighAgency.com a thing worth camel-casing. I believe websites are increasingly important right now. And part of Mack's genius writing is in using meme humor freely—which means effectively.

You cannot walk away from his domain without getting idea-virused. Mack brute forces the antimemetics away to incept you with his term. For those who choose to change how they live as a result, he also gives you a handle. You can thus pass on the good word to the next baton seeker you encounter.

Why the fuck does Andy write like this now? Good question. It depends.

"Agents" are today's buzzword within AI. But the problem lies in thinking that "within AI" means anything at all. It's far better to examine the etymology of agency. This is what chatGPT would do, to effectively analyze what "AI" means.

One finds YCombinator is the best institution at present that injects generalized agency into humans at scale. People have tried to conceptualize ways to package Paul Graham. Short of $500k for 7% of your business, frameworks like "ask five why's" might eventually get one there.

I'd emphasize trying 55555 why's ...heck, why not?

Certainly, one will do this after interviewing and getting rejected. But you will eventually find, like Naval and Peter, that PG turns out to be difficult to compress. That's why and how he created YC in the first place!

As a compulsive information aggregator who found it best to amplify his leverage, PG only mobilizes his own agency with surgical precision. His real stack is not Lisp, it's mostly money and language itself.

Words matter. A lot more than we used to give them credit for.

Coming back to questions, let us recognize how "within AI" might as well be framed as "within the universe" or "within ourselves" because epistemology itself is being overwritten every day.

Humans used to wait tens if not hundreds of years to absorb new changes to society. How we interact, the kinds of questions we can ask ourselves and others, such relationships evolved with frustratingly graceful turtle speed.

But the hare is the new tortoise. Learning systems are profoundly disturbing to standard notions of reality.

Any learning system under optimization will find the minimum description of the task’s invariance structure.

That is not just true of AI. It is true of a child learning language. A musician learning to improvise. A scientist building a theory. A culture developing myth.

Intelligence at every scale is one thing: finding the underlying rule.

The moment when a learning system finds the minimum description of its task is the moment when you don't need to ask any more questions.

Trust me, I've tried answering lots of questions. Mostly because I'm always asking.

My questions may not look like questions. Every action one takes is a question. "What happens next?" we constantly wonder as we do things, if we're thinking deeply enough to be cognizant of the moment and the action we are engaged in.

People say "there are no stupid questions" but the smartest people generously say "here was the question you meant to ask."

The zen koan approach compresses this further. Yes, you got it...

"It depends."