Bromeo

openrouter->rankings shows a pareto frontier. https://openrouter.ai/rankings#benchmarks

DeepSeek claims that they can run inference on Huawei chips. Not sure about training.

How does it compare to opensnitch? https://github.com/evilsocket/opensnitch

This works with the openrouter API as well, which skips having to make a google account etc. Here's a Claude-coded openrouter compatible adaptation which seems to work fine: https://github.com/RomeoV/gemimg

A 1024x1024 image seems to cost about 3ct to generate.

Are you familiar with the "Bitter Lesson" by recent Turing Award winner Rich Sutton? http://www.incompleteideas.net/IncIdeas/BitterLesson.html

One thing I'm missing when making slides with typst is the ability to show short videos or animated gifs. Although to be fair this isn't easy in beamer either.

Typst can actually include gifs, but they don't move for me. I have some hopes that perhaps one could make slides straight in html which could alleviate the issue.

Looks like the performance is pretty decent, somewhere around Llama3.1 for general knowledge (Tables 17) but still a bit behind in Code and Reasoning (Table 18). Llama3.1 was released about one year ago.

Per-user search history can directly be sold to advertisers, no? I was under the impression Google and Microsoft do that, or at least use it internally to build a profile of each user, again used for advertising.

Kagi answers that there is "CBMC", which is single-threaded, but that there are extensions "Deagle" and "Yogar-CBMC" that provide multi-threading for CBMC. It gives links to the papers for all three, however some of them are closed access (or in other words, fact checked, unlike arxiv).

I just stumbled across this commit and couldn't believe my eyes, thought it may be interesting to some of you as well.

EDIT: Looking further into it, it looks like there's almost two thousand commits on github starting with this exact hash: https://github.com/search?q=hash%3A0000000&type=commits&p=1

Yup, looks like org-babel is at least 15 years old. https://github.com/taruti/org-babel/tree/master

I don't think it ever had huge adoption across whole teams, but I hope if there are new implementations that they take away a number of lessons you can gather from 15 years of org-babel.

I suppose it's "RNG" if the commit has exactly one 'e' and otherwise only numbers, so that YAML interprets it as scientific notation. I assume otherwise it's always interpreted as a String, as a fallback.

I find this comment to be very tone-deaf. Are you interested in buying stocks to support Venezuelan companies, or are you purely trying to make a quick buck off of the backs of Venezuelan people trying to prevent their country from being a dictatorship?

Very interesting. The abstract claims that although GPT-4 was claimed to score in the 92nd percentile on the bar exam, when correcting for a bunch of things they find that these results are overinflated, and that it only scores in the 15th percentile specifically on essays when compared to only people that passed the bar.

That still does put it into bar-passing territory, though, since it still scores better than about one sixth of the people that passed the exam.

I really don't mind the green ones, in particular when the bikes are used as an alternative to a second car, which to my is significantly more "fugly".

Thanks for the nice writeup! So how many hours does your laptop get after all this?

It would also be interesting to see an "all default" install (arch or manjaro) somewhere, is someone knows a link.

I don't want to read too much into it, but the person (supposedly) submitting the PR seems to work at 1Password since December last year, as per his Linkedin. (And his Linkedin page has a link to the Github profile that made the PR).

Probably they could get a nice speed up by changing the if condition from your link into a multiplication with (eoe != 0) two lines down instead.

Do you sync your passwords between devices?

Another nice thing is that boilerplate code can be written in a block which is not exported to html or pdf, and therefore doesn't pollute the reading experience, but is still available in the org file.

I took a course in University (Eth Zürich) which referenced "Chapters 5 and 6 in Computer Systems: A Programmer’s Perspective".

I believe you can explicitly set "overlap=false".

I was also surprised, but it seems to be true. Wikipedia lists the median disposable income after adjusting for local prices (PPP), and USA is first on that list.

https://en.wikipedia.org/wiki/Disposable_household_and_per_c...

The site has just been taken offline by Drew due to the unfortunate start. I hope we can come back to this once the project has been properly launched, although Drew notes that he is "really unhappy with how the roll-out went" and that "my motivation for this project has evaporated" [1].

Thanks for all the work Drew, I hope you guys manage to come to a conclusion that you are satisfied with!

[1] https://paste.sr.ht/~sircmpwn/048293268d4ed4254659c3cd6abe67...

Here's the github repo used to create books in the same style: https://github.com/sisl/tufte_algorithms_book

Note that this repo is from the same author as the book.

Previous discussions:

  Algorithms for Decision Making(http://algorithmsbook.com/)  
  693 points|Dowwie|1 year ago|85 comments

  Algorithms for Decision Making [pdf](https://algorithmsbook.com/files/dm.pdf)  
  498 points|mindcrime|3 months ago|50 comments

This is solved well in Wayland, with X11 this is not as straight forward. Wayland comes with some other issues though.

Unfortunately the theory for linear models does not translate easily to deep learning based models, which this demo is based on. The "strength of model fit" becomes much more complicated and is an active field of deep learning research.

I think many people commenting on the model making bad predictions are missing the point. The speaker argues that even though models are known to be inaccurate, companies like tinder or insurance companies might still use the model outputs since they have nothing better. Therefore, in some future (or already today?) you can suffer from bad model predictions because you are "not normal enough" for the model to make good predictions, and might therefore receive a wrong predicted life expectancy and higher insurance bills.

so 2/3 for them I suppose

In Zürich, a monthly city ticket is around $50, and a monthly country wide public transport ticket is $340 (and really gets you almost everywhere). Annual tickets are slightly cheaper.

I would argue while not "very cheap", the prices are still reasonably cheap, especially when compared to reoccurring car costs.