eyegor

Correlation != causation. Especially when the study doesn't show a simple correlation like that either. At least not for heart attacks.

> The multivariate survey‐adjusted logistic regression analysis of the entire sample did not reach a significant association between daily cannabis use and CHD (adjusted odds ratio [aOR], 1.16 [95% CI, 0.98–1.38]; P=0.09)

But there does seem to be one for strokes

> Among those who had never used tobacco cigarettes and had never used e‐cigarettes, cannabis use was not associated with CHD or MI but was associated with stroke (aOR, 2.24 [95% CI, 1.31–3.83]) versus tobacco‐adjusted odds in the whole sample (aOR, 1.42 [95% CI, 1.20–1.68])

Study https://www.ahajournals.org/doi/10.1161/JAHA.123.030178

A major political party and several large corporations are opposed to increasing federal funds to the post office. Unless you can spin it as national defense, notable federal funding for any group is unlikely.

Do you mean cython? I've never seen a pure c extension in "a few lines", last time I wrote one there was a ton of boilerplate.

Eh, it's uncoated low chrome stainless (some 300 blend). It wouldn't take much to start surface rust. In my experience low chrome stainless will rust in months if it hangs out too close to normal steel (some kind of galvanic migration?). Since they just started delivery and it's winter, you'll know for sure (by spring) whether it was contamination or if they're all doomed. Road salt is much worse than salty air.

I suspect you meant to post this elsewhere... Otherwise I'm missing the link between rusting trucks and the food supply.

Wait I still do this, what are your secrets?

Is there any compiled language that doesn't benefit heavily from release builds? That would be interesting if true.

To your last point, have you tried plotnine? It's meant to be ggplot2 for python.

https://github.com/has2k1/plotnine

Why do the first dozen examples not have screenshots in the main readme? I thought maybe it was an issue on my end until I checked the raw view.

For $1000 you could basically build one today. 5950x (16c cpu) is $400, 128gb kit of ddr4 is $250-300, maybe enough budget left for mobo, ssd, psu, case, cooler if you bought used bits (may also need cheapest gpu on Craigslist depending on your needs). All new ~$1200. Aws/gcp prices are nuts compared to on prem, although hetzner and co would be much cheaper.

Riddle me this, where did the magic 132 columns come from? I want to say it was fortran77 when extend source was added to let fortran files use 132 columns.

See also https://scottplot.net/ although it's not avalonia specific, it just has an avalonia option.

I see the potential long term but that api is unintuitive at best. It almost seems obstructively terse.

    (ggplot(mtcars, aes("wt", "mpg", color="factor(gear)"))
     + geom_point()
     + stat_smooth(method="lm")
     + facet_wrap("~gear"))

From https://github.com/has2k1/plotnine, their front page examples. I get that it's ripping off the ggplot api, but why would they not alias "aes" to something more meaningful? It stands for "aesthetic mapping".

My problem with all the popular plotting libraries in python is they seem designed for researchers slapping together a paper and produce code that's awful to maintain.

Besides pointing pentester tools like metasploit at yourself, there are some nice scanners out there. Examples in no particular order:

https://github.com/quay/clair

https://github.com/anchore/grype/

https://github.com/eliasgranderubio/dagda/

https://github.com/aquasecurity/trivy

So then you set up something like a cron job to scan everything for you and email the results once a week or whatever if you don't want to monitor things actively.

Numerics code is annoying to maintain, and most of us have code bases that already use mkl. Intels compilers make it easy and quick to use, whereas openblas was always a headache in my experience. Plus amd hardware only became competitive in the past few years, so it hasn't mattered.

Holy cow, my main takeaway is that mkl is complete garbage on amd parts and that isn't on accident. This makes me think I need to go back and relink some programs against blis or openblas for amd systems... Even though we primarily switched cluster hardware to amd it never occurred to us that mkl would run at half the throughput of open implementations.

Meanwhile sleep() in js land is more of a suggestion when it comes to accuracy. Browser standards are strange.

Of course I mean setTimeout/interval because js doesn't even expose a sleep function.

I run tabby [0] which uses llama.cpp under the hood and they ship a vscode extension [1]. Going above 1.3b, I find the latency too distracting (but the highest end gpu I have nearby is some 16gb rtx quadro card that's a couple years old, and usually I'm running a consumer 8gb card instead).

[0] https://tabby.tabbyml.com/

[1] https://marketplace.visualstudio.com/items?itemName=TabbyML....

The 1.3b model is amazing for real time code complete, it's fast enough to be a better intellisense.

Another model you should try is magicoder 6.7b ds (based on deepseek coder). After playing with it for a couple weeks, I think it gives slightly better results than the equivalent deepseek model.

Repo https://github.com/ise-uiuc/magicoder

Models https://huggingface.co/models?search=Magicoder-s-ds

I'm more curious what led you to contact a lawyer over a $20 loss... I don't think I've met a lawyer who was cheap enough for that to make sense. If you're going to fight a vendor, why not just do a bank chargeback?

Unfortunately, door 1 is maybe $200 bounty and weeks or months of back and forth (if the corp doesn't have a clear bounty program) whereas door 2 has infinite upside. Honestly, it might make sense for a gov group to run a standardized bounty program for exploits with notable financial / privacy impact.

In my experience, pypy works with basically everything these days. I remember having some struggles with a weird fortran based extension module a few years ago, but it might work now too.

Most c extension modules should work in pypy, there's just a performance hit depending on how they're built (cffi is the most compatible).

https://doc.pypy.org/en/latest/faq.html#do-c-extension-modul...

Lmstudio (that they linked) is definitely not open source, and doesn't even offer a pricing model for business use.

Llmstudio is, but I suspect that was a typo in their comment. https://github.com/TensorOpsAI/LLMStudio

So you don't see significantly worse performance on 3bit quantized models compared to 4? Every 7/13b model I tried quantized gave much worse responses at 3 bit and below, whereas the differences from 4 bit to 6 or even 8 bit is more subtle.

It's wild to me that they would write performance focused code (unsafe ptrs, memory mapped files, etc) and then still use linq in the main work loop [0]. This implementation could go much much faster. Of course rewriting the parallelism and aggregation from linq is a decent amount of effort, but I'm betting that applying the lazy man's hack [1] would have decent gains.

[0] https://github.com/buybackoff/1brc/blob/main/1brc/App.cs#L12...

[1] https://github.com/reegeek/StructLinq

Probably not 5x, more like 5-6k. The a6000 is $7k and this is a cut down version.

I tried kagi for a bit, but these days I think phind gives the best results for cs related searches. In addition to good search results, it shows llm responses with a chat interface to continue that chain.

https://www.phind.com/

As an alternative you could write some automation scripts to handle all the requirements for self hosted install. If you look at oogaboogas text-generation-webui [0] you can see what I mean. Although ease of install also leads to a larger user base, which can be a double edged sword for something as ubiquitous as a pdf app. It's much easier to get people to submit issues than to help solve them.

[0] eg, windows install script https://github.com/oobabooga/text-generation-webui/blob/main...

pdftk has an issue where it corrupts pdfs occasionally in modern windows server versions. Had a weird bug in a random helper service at work that we narrowed down to pdftk mangling documents sometimes. Never looked much into it since it was only a couple of hours to replace it with another tool, and haven't had issues since. I think all we used it for was merging and adding watermark text.

As a long time user of intels scientific compiler/accelerator stack, I'm not sure if I'd call it a "good track record". Once you get all their libs working they tend to be fairly well optimized but they're always a huge hassle to install, configure, and distribute. And when I say a hassle to install, I'm talking about hours to run their installers.

They have a track record of taking open projects, adding proprietary extensions, and then requiring those extensions to work with other tools. This sounds fine, but they are very slow/never update the base libs. From version to version they'll muck with deep dependencies, sometimes they'll even ship different rules on different platforms (I dare you to try to static link openmp in a recent version of oneapi targeting windows). If you ship a few tools (let's say A and B) that use the same dynamic lib, it's a royal pain to make sure they don't conflict with each other if you update software A but not B. Ranting about consumer junk aside, their cluster focused tooling on Linux tends to be quite good, especially compared to amd.