Discussion about this post

User's avatar
Adam Shimi's avatar

As someone who benefited enormously from you bringing back this advice over and over again, I wanted to add something about the self-reflection use of this.

My experience has been that the main way in which this injunction feels wrong is that I think "I know the obvious thing is not going to work".

First, if you have never done it before, how do you know that?

* If it's only from your own beliefs and models, as Gabe says above, this smells bad and is highly susceptible to post-hoc rationalization of not doing annoying/painful things.

* If it's from other people/experts, then we're back to the main point of the post: check if they have done the obvious thing, otherwise you're delegating modeling to people just as ungrounded as you.

But there is another reason: even if you were right that the obvious thing is not going to work as is, doing it is probably still your best move. Because what you cannot know without having done the obvious thing is where and why it fails. And since the obvious thing is generally simple, naive, cheap, this gives you feedback from reality much faster than more advanced plans.

(Espectially since a classic failure mode of Galaxy Braining is to never start, since you can always find some weird edge case that you absolutely need to patch before even trying anything...)

Expand full comment

No posts