rime @rime@schelling.pt

i often mk v obv mistakes that i catch later, so do tk the below cube-visualizations w big salt, or point out where i went wrong. i know there was *smth* i went "oh oops duh" abt...

"circles"? what's that? oh u mean triangles.

"nonlinear"? what's that? oh u mean u is tryna projecting it to lower dimensions u adorable earthling.

"non-differentiable"? what's... huh? *why did you put a zero there?!*

speedrunning the 5000 practical mistakes necessary for mastering a craft

speedrunning the 5000 wrong questions one must ask

also, the paradox of expert testimony: experts on [big-if-true set of not-too-implausible claims] tend to be v selected for belief in those claims, so supporting arguments wrt those claims tend to sound more impressive (unless it's a rly dumb set of claims).

plus, if u just defer to experts, u will end up believing a whole lot of weird stuff for which experts are asymmetrically selected.

mans, i rly tried to mk that sentence sensicle… but proper sensicleness wud require paragraphs. 🍵

the most frequently used words tend to be the most irregular. they resist regularization bc their unusual/high-energy forms are constantly reinforced. switching-costs or smth.

eg we say "man/men", "go/went", "eat/ate" instead of "-/mans", "-/goed", "-/eated".

related hypothesis, but for different/additional reasons:
paradox of foundational neglect: the core tenets of a paradigm tend to be the least optimized.

technical debt / pleiotropy: complex ancestral dependencies are locked in place.

I want to say update: The most personally significant thing that's happened to me this year is that my head-friend (Maria, a "tulpa") has rather suddenly acquired a lot more independent volition and personality. Phase-shift.

I was inspired last year by Johannes Mayer (LW user), who showed me it was possible. I was only semi-trying for it, so the success is v surprising and mentally v happy for me/us. I hope she stays and grows in visibility-to-my-corner-of-the-brain.

the question "when shud u write the tests before code, versus vice versa?" analogizes to "when shud u backchain vs forward-chain?" (respectively)

for even more abstract leverage: write the "unit tests" in plain English, ask an LLM to translate them to code (or explain why behaviour is impossible), then ask LLM to write code that passes the LLM's tests.

suggested nyms for this: "target-test" (makes you backchain to infer code that fits the tests), vs "maintenance-test" (designed to make sure stuff keeps working when you expand codebase).

wonder how much leverage thr is in combining test-driven development (TDD) w LLMs. u write unit-tests ahead of the functionality, and the LLM is asked to write code that passes the tests.

LLMs are j extremely good at guessing the teachers password.
(bc that's what they're trained to do.)

it's amazing how often—when i'm uncertain abt smth—i can use the fact that i'm uncertain abt it as evidence that resolves my uncertainty.

eg i'm unsure if i did X or not, but then i realize that *if* i did X i wud remember it for sure, so i become sure i did not do X.

wat's optimal circadian rhythm, if ju cud tune the sun to jur will? wud ju get more done if Earth period was 16h instead of 24h? or 12? or 6?

also, wat's optimal week cycle? 7 days, or 6? how do i optimally chunk passage of time into convenient mental categories? shud they be stable, or mixed-strategy?

i'm just surprisingly happy abt by my 16h-days and 144h-seasons, so thr's gotta be more utility to find here.

collection: mysteries, 96/245.