Endo

[Endo @hakuhobr]

@DevanshuXi @Docker Make the environment reproducible with nix then run plain binaries (eg process-compose), no need to run containers most of the time

[Endo @hakuhobr]

@thdxr If you’re using nvidia then the driver makes a huge difference, up until 565 it felt very buggy to me, more recently less so

I did get some bugs with sway as well, but more like random stuff (segmentation fault when some application sent a message to its ipc port)

[Endo @hakuhobr]

@elonmusk @cb_doge StarCraft (brood war) is much better

[Endo @hakuhobr]

@chesscom > To solve chess, you'd have to map them all

No? Checkers is solved and not all positions are mapped (which yes would have been impossible in this planet’s lifetime)

[Endo @hakuhobr]

@coproduto Verdade quanto ao tamanho da tabela, mas acesso a recursos é questão de autorização e deve-se assumir que qualquer ID se tornará público

[Endo @hakuhobr]

@ChShersh I was about to post the output of du -sh /nix but it’s been running for 5 minutes

[Endo @hakuhobr]

@ibuildthecloud Shouldn’t :)

[Endo @hakuhobr]

@ibuildthecloud That’s what sops is for, and in particular most devs should have access to production services through env files

[Endo @hakuhobr]

@doppelbanger__ Because you might press restart instead of power off and then we’re cooked

[Endo @hakuhobr]

Because at the limit there isn’t much useful to do

I think the most realistic (optimistic) scenario is (1) entertainment/social and (2) purely competitive endeavors, just like how like watching people play chess for fun, even though engines are much better, I think all kinds of competition that we’d find pointless nowadays are going to appear

[Endo @hakuhobr]

@infraflakes Chill, man. That’s my perspective, that’s fine if you disagree but your reaction makes you sound mentally unstable

[Endo @hakuhobr]

@infraflakes By “leeway” I meant that a suboptimal implementation somewhere is less likely to be disastrous

[Endo @hakuhobr]

> If you approach a language decision with how much slow code it allows you to write

That’s your own distorted interpretation of what I said. When one has limited time, using a language that gives better baseline performance is desirable because one may then use time that would have been spent on basic optimizations on other issues (features, deeper optimizations…)

[Endo @hakuhobr]

If you take the same architecture and implement in Python vs C, generally the baseline performance in C will be better before any optimizations. If you end up making a mistake (eg over processing some data or whatever), the hit in your program will be felt more harshly in Python than C. And so on

[Endo @hakuhobr]

CS papers tend to overuse math formalization/notation, which sometimes makes it harder to understand a simple concept (eg to describe something as simple as “clear least significant bits” they’d introduce a field, use a modulo, etc). In some subfields this seems on purpose as some conferences will reject papers that look too simple

[Endo @hakuhobr]

@laplacian28 It’s a ridiculous system. You can end up paying a lot less if your yield is high (because no capital gains tax), but then you have to hope the tax authority doesn’t deem those gains “business activity” and then you’re liable for the higher rate anyway

[Endo @hakuhobr]

@ShaanVP That’s more than enough knowledge

[Endo @hakuhobr]

@ilyasut Very likely already did

[Endo @hakuhobr]

What it sounds like when people say the issue with the OAI-HF incident was the sandboxing

[Endo @hakuhobr]

“Of course the international master lost to Stockfish, we just have to put proper grandmasters to play against it”