niplav @niplav@schelling.pt

When running an RCT on yourself, you still get half the benefit in expectation¹, but you also get the value of information, so it should be surprisingly cheap

¹: at least, more if placebo effect applies

Is "chatbots are mostly used by women" good or bad for daygame?

This toot was written on the unceded lands of the usenettites

Oh look this alread, exists

https://en.m.wikipedia.org/wiki/Fractional-order_system

Thoughts had while thinking about the dynamics of Omohundro sludge, where there's meta-dynamics of trying to outwit on higher orders all the time, while keeping in mind all the lower orders of the game

Also related to fix-point solutions for regress problems

Transfinite-order differential equations

Might sometimes result in differential equations on smooth functions where one can take the infinitieth derivative of a function

Differential equations where one variable also determines which derivative of a function is used

f^{(k)}(x) = log_{x}(k)

Using fractional calculus of course

I wonder what the simulators think about my recent life decisions

Demon Core Transformation

Rationalist honesty norms are trauma-caused

Llama model distribution is commodification of complement if I've ever seen one

Post: Precision of Sets of Forecasts

niplav.site/precision

I conclude that methods based on rounding probabilities are hot flaming garbage, methods based on rounding odds or log-odds are regular garbage, and methods based on applying noise to log-oddsare ok but sort of slow.

Midnight is the best time to fire off a bunch of aversive emails