jeremyjh

11 ms·

181.
by jeremyjh 3mo ago
Yes and at 1-ply Leela will destroy Stockfish because her eval is so much better. Searching is slow even when its really fast.
182.
by jeremyjh 3mo ago
Step one means fuck-all nothing with a cap of $800K. Its not as if billionaires truly pay $0 in tax. $800K is approximately $0 to a billionaire so they'd all just pay that.
183.
by jeremyjh 3mo ago
Yes and they don't mention any language at all for Windows. Just "native", which is pretty much a philosophical question for that platform. I guess that means the founders have all avoided that platform their entire lives - w
184.
by jeremyjh 3mo ago
That's probably not the type of work/life balance they're looking for employees to have.
185.
by jeremyjh 3mo ago
"I completed the Frobnicator challenge" probably works just as well. No one would want to admit they'd never heard of it.
186.
by jeremyjh 3mo ago
Without clicking you’d never know why it’s the title.
187.
by jeremyjh 3mo ago
libghostty exports a C ABI. For an application from Mitch at this layer I would guess Go.
188.
by jeremyjh 3mo ago
I like the bloom filter idea - I agree I don't want every URL I visit to be sent to Algolia. I really like this project's UI though - automatically opening it in a pane so I don't have to leave the site to see comments is rea
189.
by jeremyjh 3mo ago
It can easily be both. Also, they didn't use this innovation in K3 - K3 pre-training would have started months ago and the paper only mentions a 48B model. The people working at this level may not even be heavily involved in shipping a
190.
by jeremyjh 3mo ago
Well then, nevermind.
191.
by jeremyjh 3mo ago
Whats an example of magic in the meson.build that you don't understand? I think this is the first meson.build I've ever looked at and it seems very straightforward.
192.
by jeremyjh 3mo ago
I just looked at one example...Wayland's meson.build is 142 LOC, but build.zig is 581. https://github.com/allyourcodebase/wayland/blob/master/build... https://github.com/wayland-mirr
193.
by jeremyjh 3mo ago
> indicating that all of the issues that justified the rewrite were self-inflicted and addressable The main issue that justified the rewrite were memory bugs, especially related to interaction with GC managed Javascript objects. There i
194.
by jeremyjh 3mo ago
The X are like comments, you set those when you've determined a square cannot have a sunflower on it. If you made those double-click and flowers single-click, it would be easy to make a mistake like not double-clicking fast enough and
195.
by jeremyjh 3mo ago
That isn't really relevant to GP's point. We still don't know what training costs because we don't know how much RL is done.
196.
by jeremyjh 3mo ago
No, its the website's choice. They want to track you. The government said they need permission to do that. They could offer a better experience by just not tracking you.
197.
by jeremyjh 3mo ago
They don't love you. They want to track you and make a tiny amount of money off that. Also, they want to hire cheap incompetent staff who don't understand what all the analytics shit they are putting on the website even does.
198.
by jeremyjh 3mo ago
This is the only way. They'd also have to make all third-party assets illegal as well though and that may break some legit use cases, and make it a lot harder to use CDNs.
199.
by jeremyjh 3mo ago
The companies want to track everything you do on the web. EU passes a law that says they have to ask for permission first. They all comply maliciously because they still just want to track you. Then you blame the government.
200.
by jeremyjh 3mo ago
Putting assets on a 3rd party server is indeed a great way to track people. They can completely fuck off with all of that as well.
201.
by jeremyjh 3mo ago
The entire banner is malicious. They don't need consent for necessary or functional cookies. They only need consent to track you - at no benefit to you ever.
202.
by jeremyjh 3mo ago
Yes we agree with each other but the objection you were replying to is laboring under the misconception I described - it’s not a semantic distinction.
203.
by jeremyjh 3mo ago
It has nothing to do with thinking or consciousness. There is a common misconception that LLM are simply a "statistical process" that doesn't feature any abstract conception of the tokens it is predicting. There are studies t
204.
by jeremyjh 3mo ago
The statistical model that underpins any deep learning system is the substrate in which a process is implemented. There is still a process - just because that process is grown - not programmed - doesn't tell us anything about the dep
205.
by jeremyjh 3mo ago
You just stated your vibes like it is a fact. There was a study about this: https://journals.sagepub.com/doi/abs/10.1177/095679762412430...
206.
by jeremyjh 3mo ago
Microsoft has many of the same advantages but is incapable of judging the quality of the products it makes.
207.
by jeremyjh 3mo ago
It assumes git works the way that git works, and that git will not break backwards compatibility in the future. I think the same can be said of any alias in git.
208.
by jeremyjh 3mo ago
I would never do this as a visitor to another country - but if this happened where I lived I would advocate mass littering until it was repealed. Just drop the trash where the canvases to be. If you could get even a single percent populatio
209.
by jeremyjh 3mo ago
This entire article is about the fact that the codebase deteriorates over time when you work this way. If they are correct, it will take longer and longer to develop PRs, and more and more cycles to QA them. Eventually the code will be unsa
210.
by jeremyjh 3mo ago
Is that why Linux failed? Just imagine the world in which an open source kernel dominated almost every compute market on the planet!

More ›