memoriyato3

standard deviation is misleading for non-standard distributions (fat-tailed, skewed, multi-modal, ...)

common mistake people make

so the proof to the unit-distance problem was on the manifold, given it was outputed by a LLM?

is the proof to the Riehmann hypothesis also somewhere on the manifold and we just need to prod the LLM with the right prompt so it locates the point?

Ironic that this website has exactly the regression to the mean kind of visual design that LLMs output

well, Google refused to increase Meta quote of tokens, even Google can't supply so many (paid) tokens as Meta is burning

I use an app to switch the Windows system theme between light/dark automatically based on time of day (similar to auto blue color reduction).

It's funny noticing how most Electron/WebViews/web-sites immediately switch too, and have good dark mode support, while non-web-tech native apps either only support light-mode, have a bad looking incomplete dark-mode, or require a restart to switch.

So much for "native GUIs are superior, consistent and respect the user". Microsoft is still struggling with adding dark mode support to most Windows included apps.

that's like "stop using a car, you can walk between cities and its healthier"

programmers were always against "software patents" - the idea of copyrighting algorithms and implementations

it was impossible to write code for 8 hours straight, you naturally had to stop

but you can prompt for 8 hours more or less

like running versus cycling - you can cover more distance by cycling and it's less intensive (I'm talking casual running/cycling, not racing)

agents are a bicycle for the mind

I have enabled "charge up to maximum 80%" setting on my phone from day 1, and 2 years later the battery lasts noticeably less. maybe it would have been even worse without that setting.

its called hedonic adaptation - you get excited by a new model, but then the excitement disappears, and you confuse that with the model being nerfed

having E2E encryption is a marketing feature, you need it if you want to be competitive in the market, so this is another incentive to add it

technically you could audit your local copy of tor source code, build it, and then never upgrade it.

still this wouldn't guarantee that all the other nodes are not compromised

we have EVs for almost 15 years now, one would imagine some real world data on battery life was collected already, even if not widely shared

the propaganda works because we are all familiar with the smartphone battery degradation after 2 years, so it seems logical that same thing will happen with car batteries

psychohistory, some are already doing it today (LLMs simulating various people with various beliefs interacting)

you can still judge if it was correct or not, irrespective of contrarianness, still a good signal to measure

to quote Taleb: those who can, do (trade). those who can't, write about it (financial press)

also helps adding negative conditions like "do not nitpick", or specific bad attractors that you see "do not investigate/report anything related to symlinks, they are not a concern"

yeah, happened to me: "A is very wrong, you should do B", and on the next fresh review loop "B is very wrong, you should do A"

typically this means there is some ambiguity in the specification, and the model flips between alternative interpretations

for SecureBoot you could use the Linux shim bootloader, to boot your stick, or a tiny Linux that runs your code, right?

on AMD/Intel CPUs from the last 5 years you just need to enable hardware memory encryption in the BIOS