Borges' cartographers and the tacit skill of reading LM output

https://galsapir.github.io/sparse-thoughts/2026/04/11/map-and-territory/

by galsapir • 3 months ago

40 10 3 months ago

galsapir.github.io

gaythread• 3 months ago

From the bio: i’m a researcher (working in healthcare). I try to consume content — books, papers, posts, podcasts, whatever — and started writing to figure out what I actually think about it all. This is writing from the edge of understanding. Hopefully, less about “here’s what I know” and more about “here’s what I’m trying to work out.” Rough edges included.

contraposit• 3 months ago

https://www.goodreads.com/book/show/42041926-the-scout-minds...

This is my favourite book. I have followed their podcast for so long. You might like the ideas in this book. Or you can watch the 1-hour lecture by her on youtube. We needs scouts to faithfully explore the territory and report back.

https://rationallyspeakingpodcast.org/ by Julia Galef and Massimo Pigliucci

galsapir4• 3 months ago

Thanks! I'll check it out.

operatingthetan• 3 months ago

Not going to read something where the author can't do the basic courtesy to the reader of using capitalization.

galsapir4• 3 months ago

haha that's a style choice (takes more work to get lowercase text these days). But yeah legit ;-)

RealityVoid• 3 months ago

I found it more peculiar that "LM" was capitalized. I know the style is very trendy in certain circles right now, but it is a bit puzzling for me.

nagaiaida• 3 months ago

that is your prerogative, but this seems to me far more discourteous than somebody putting information down in a format you're perfectly capable of reading yet dislike.

tyrust• 3 months ago

Nice piece.

Right now the skills you describe are definitely relevant. At work I'm regularly reviewing smelly changes, both from my own agents and others'. I'm wonder if this smell will always be present or if it will go away entirely, leaving the smell detector skills irrelevant.

galsapir4• 3 months ago

Hey thanks! I do wonder that. I think that even if specifically for code smell the things would be subtler, for other forms of AI driven averageness (especially in areas where we can't RLVR the models to perfection) it might still be present. But yeah I wonder how those thoughts will age (and how we'll update our priors accordingly).

Borges' cartographers and the tacit skill of reading LM output

Related Stories