valyagolev

5 ms·

1.
The game of semantic promises (Or, why you are not using Siri)(salamilang.substack.com)
1 points by valyagolev 2y ago|0 comments
2.
Slalom: Designing a runtime for AI-generated code(slalomlang.substack.com)
8 points by valyagolev 3y ago|3 comments
3.
by valyagolev 3y ago
it'd be a fascinating feature - a random number on the face of the watch - especially on a mechanical one
4.
by valyagolev 3y ago
perhaps because it gives the same answer, verbatim, for many different attempts to figure it out? and because we know it well enough to be sure that it's not smart and devious enough (yet) to conspire like this? (nor has any clear reas
5.
by valyagolev 3y ago
After discovering Muji Gel pens a decade ago I just buy them tens at a time (pens or refill sticks). They feel absolutely perfect and are very easy to obtain in Europe. It's a great feeling to find a product that works so well you can
6.
by valyagolev 3y ago
we should code up a real online game to gather the stats here. maybe we'll find a real subliminal messaging model
7.
by valyagolev 3y ago
so it hallucinates when explaining what these companies do in relation to an idea. basically it's an automatic downer machine that lies that there's a lot of companies making your idea
8.
by valyagolev 3y ago
you also should, and inevitably will, cut corners smaller than your "step" when actually following the "coastline" in development. the more fixated one on the smallest details, the longer it will end up
9.
by valyagolev 3y ago
Weirdly I'm very used to this kind of stuff and take it very easily from programmers, but when a german bureaucrat does this same thing to me I'm also fuming.
10.
by valyagolev 3y ago
there're no sanity checks because there are no sane "recovery" nor there are much risk. useless sanity checks mostly work towards making software annoying and unusable
11.
by valyagolev 3y ago
Not engage on sensitive topics: yes, even with people you like Not engage on any topic outside the strictly necessary for work: a passive-aggressive way to keep the animosity forever
12.
by valyagolev 3y ago
they always insist that Leibnizian monad is not related to the FP monad at all, probably because it's a major point of pride of contemporary scientist to not be related to anything philosophical at all. sure. still, i'm always won
13.
by valyagolev 3y ago
even simply expanding consultants' liability is bound to help them: they have more capital and more lawyers to fight, while any small shop would be destroyed by the first liability-related lawsuit
14.
by valyagolev 3y ago
> with consulting there are no certifications, no licensing — it’s kind of a Wild West mentality I imagine how bad it would become if you'd have to get an official certification to do consulting work
15.
by valyagolev 3y ago
it is quite sensible that origin is top left in potentially scrollable situations
16.
by valyagolev 3y ago
Add Cosmoteer, where you build and fly a starship: playing it makes me think of John Boyd
17.
by valyagolev 3y ago
Borges: "There are only four stories: the siege of the city, the return home, the quest, and the sacrifice of a God"
18.
by valyagolev 3y ago
the man has a whole career out of pushing corporate-controlled languages into wider and more open use
19.
by valyagolev 3y ago
tobacco patches do it for me
20.
by valyagolev 3y ago
I wonder if there's a subsubsubfield of ML research that goes like "let's use halting-evaluator as an activation function for our purely theoretical neural network". or like, "NNs with oracle"...
21.
by valyagolev 3y ago
Parkinson's Law is probably the same thing, but framed in much more hostile way (now it's the worker being lazy/creating extra work/yak shaving...)
22.
by valyagolev 3y ago
I'm not sure that's an inevitable conjecture. It's not the original text samples from the web that govern how chatgpt acts in chat - it's the consequent chat-specific fine-tuning. Also I'd be surprised if anxious t
23.
by valyagolev 3y ago
yeah when GitHub is sold to PE 15 years from now, and we get emails like "you have 2 days to backup all your repos", i hope I won't feel too lazy
24.
by valyagolev 3y ago
If you want controls and scrolling for the youtube short link, just use the short id as the video id in the normal youtube url: https://www.youtube.com/watch?v=ePvpYVNEvzk
25.
by valyagolev 3y ago
I'd rather love to have a truly ergonomic controller with full shoulder width
26.
by valyagolev 3y ago
Never thought about it this way, but yeah makes a lot of sense. So the real approach should be: pick the kind of institution you care about first. Skill mastery might just end up being optional. To be honest, I'm rather surprised how g
27.
by valyagolev 3y ago
this happened to him several times, and each time he ended up betting it all again. if anthropic would make him whole, he'd bet it all on more bullshit crypto
28.
by valyagolev 3y ago
Excited for the next 100 years for this AI-related equivalent of "you got the wrong result, because your query is slightly not exactly the way system expects it to be". ChatGPT knows a good answer to the question even without embe
29.
by valyagolev 3y ago
i don't even need it to go beyond the insides of a function. something like a proof of an invariant that's only relevant inside the function's body would be fine. e.g. in Rust, almost every place where you see something like
30.
by valyagolev 3y ago
I asked this one about homosexuality, it didn't find the most glaring passages from Leviticus. This is a common thing for vector similarity search. I wonder if there's a solution already. I thought about giving the query to an LLM

More ›