freehorse

I puzzled over the "The Long and Short of It" as for me (git version 2.54.0)

  git -h branch

  git branch --help

I see that `git -h branch`, `git branch --help` and `git --help branch` give the long output but `git branch -h` gives the short. I suspect that `git [-h | --help] branch` is converted to `git help branch` which runs `help` with argument `branch`. But `git branch -h` runs `branch` with argument `-h`. Also as `git -h alias` prints the alias definition while `git alias -h` gives the short help for the aliased command.

[0] https://news.ycombinator.com/item?id=5512103

Taking creatine does not bar you from doing other things. Being "low energy" is too general a thing, and to approach you probably need to start eliminating factors anyway until you find the issue, by getting better sleep, doing blood tests etc. Taking creatine can be part of this.

Creatine supplementation not having significant effect in X domain in population level does not mean that it will not have effect on YOU if for example you are vegan and don't get enough creatine from diet. If you have low energy, probably you have something that may not show in a population level. Similarly you may be low in B12 and then B12 can help you, but it may not show effects in population level if most of the population is not low in B12 (and your study is not powered enough to catch nuances).

It is not like when you have to choose between two medicines and excluding one can be harmful even if the other is not harmful to take per se. You can supplement creatine and try whatever else at the same time.

What are the "downsides"? It is one of the most studied substances and except in special cases like kidney issues it is totally safe.

It depends on how often you search to find things, how much chronically online you are etc. I do 800-1200 searches per month, and that's excluding !w etc bangs that I also heavily use.

That's not generally true. From [0]

Across most states, viewing CSEM alone is generally not a mandated-reporting trigger; reporting becomes obligatory when disclosures involve an identifiable child being abused or used to produce material.

California’s CANRA imposes a distinct duty to report electronic access (download/stream) with identifying patient information, upheld against privacy challenges based on compelling state interest.

[0] https://www.psychiatrictimes.com/view/mandatory-reporting-ch...

Industrial espionage is neither a chinese invention nor (specifically) chinese practice. American companies do it to each other even.

But one can already switch models in a harness between llm providers. I doubt such a pivot can get some much traction.

You can do both, until you can't, and then usually making money trumps social progress.

For now, it seems that direction that lm studio is taking for enterprise market is "local ai deployment support". They recently launched "lm link" which basically uses tailscale to create e2ee connections between computers running lm studio where you are logged in. Granted one can also setup tailscale or their own vpn themselves and use llama-server, but I guess it is simpler to provide it out of the box. In any case I am not sure pivoting from running local models to "cloud offering" (as in providing llm inference at their severs) is a sensible choice granted there is already competition in that space and they have no leverage there. The highest expected path imo would be to be bought by a company that makes (esp open weights) llms and has a similar business plan around enterprise contracts with local deployments.

Maybe. But also imo it happened because the author was not testing proton/crossover at all until the beta was released, and even then only later that did people started reporting it. Developing in one platform and then post-release fixing bugs in others that the game is not actually meant to be played is probably not easy.

The bug is still there, disabling assisted moves helps a lot and it triggers less but it still happens if you are unlucky.

Just to note that the author seems in general very nice and helpful, and one can choose what platforms they want to support, I just wish we could get native builds for more platforms like with duskers 1, which had native macos and linux, or at very least to have wine supported so that there is less of windows monopoly in games.

Author's latest title has enormous memory leaks in linux through proton (and in macos through crossover) [0]. I hope the author puts effort in non-windows platforms for this one during development, because the first duskers was indeed a great game.

[0] https://steamcommunity.com/app/1883920/discussions/0/6897422...

Not saying this device is not cool, but one can also get this info easily in a computer, if you find out what to look for. One had presented a utility here some time ago with a menu bar icon showing this information

https://news.ycombinator.com/item?id=47972511

Personally I have stopped buying cables that do not disclose this information. There are pretty fine alternatives that do, so I see no reason to take gambles.

My only issues so far come from charging protocols rather than cables anyway.

Moreover, stuff like how many watts a cable supports are issues that happen regardless connector type.

Blogspam and corporate blogs had killed quality bogs ages ago

One of the main reasons that I (and I assume others) am here is because I can discover interesting content. It is true that there is a lot of spam in the internet but if I wanted that I would be in x, linkedin or sth. My problem with AI right now is that I consider machine generated content low quality one, and I would like to be able to decide if I want to read such an article without having to waste time before I realise it is ai generated.

if an AI generated article is indistinguishable from a human written article and it is accurate and interesting, do you care who wrote it

Ideally I would like to know regardless. Practically in such a hypothetical future scenario it may be impossible, but I think that not doing sth right now because of some hypothetical future that may or may not happen is not a very good argument. Right now the AI content is pretty distinguishable, and if one took the time and effort to make it not seem like AI then at least that text contains some more human effort.

Typical linkedin posts always sounded like that and they were already not well liked.

Oops — sorry.

I dont see that. To me it reads like a genuine question to an actually underspecified part of how much embedded carbon emissions ram entails. I read the question at the time and found it reasonable and genuine, and I see no strong reason to suspect that any of the commenters there is an LLM.

A judgement on the content (product of writing) is not a judgement on people. I dont care if a person uses LLMs in their life – heck, I myself do. But I care if I want to spend my time reading an LLM-written/assisted article or not.

PS that's not an em-dash – this is.

The "show [dead]" setting already exists, so it is not 100% "exposed to the same feed".

AI has a distinct, low-quality and unnecessarily sensational writing style, which signals low effort post first of all. If you use AI and you put the effort to make it not look like AI it means that at least you put some effort into it.

I am not sure how the future may be, but the direction of AIs has been to solidify their distinct writing style rather than assimilate to humans. So we can start addressing what we are dealing with now. I doubt HN tagging/flagging can put enough pressure to affect how llms will evolve anyway in the future, while the llm speech becomes more and more ubiquitous and unhinged.

I mean even if you are comfortable with grep, it is still not enough, plus then you have to reinvent all the rest of the functionality, ie an extension to automatically add references from a webpage, download pdfs and keep them in a directory with some local link to be able to find and open them, and then another tool to export a subset of the huge bibtex file for each specific project and sync them, as it would not make sense to share a huge bibtex file on every latex project where you need only a small subset.

And moreover, something already mostly solved algorithmically. Eg in zotero you can drag and drop files and it generates a reference based on the metadata as long as that is enough (somebody mentions it is not possible in newer versions, I use the latest version and it works fine).

Each one can choose their own tools of course, but we often see llms shown as solutions to something presented as unsolvable before, while it already was pretty amenable to normal tools.

Isn't citation format handled entirely in the latex document itself?

Then it seems that what you want is running a repl of your chosen language rather than shell/"executable" scripts.

Sure! It is in matlab (I know...) so I assume you may want to ask an llm to translate it to python or sth but it should be fairly simple

https://pastebin.com/n1mvcxP1

What text did it find?

It is smart, but it is not impossible to crack algorithmically imo. In particular, one can take two consecutive frames and run perturbations moving one frame around (ie shift the indexes) to find where the difference between the two consecutive frames minimises. Then subtract these two (perturbed) frames and ocr it. Works easily if the movement is linear/one-directional.

I did this in 20 lines of code (checking only vertical perturbations), and this is what I get with subtracting frame 7 from frame 1:

https://imgur.com/a/only-human-can-read-this-vfDe6ZA

Maybe not so well explained, by picking the default intended text presented to be the same as the decoy text. It took me also some time to realise what was going on, but the execution is fine otherwise.

So there are two texts, one decoy (which you can barely see in a single frame but becomes more clear if you average between frames) and an actual text, which disappears in single frames or averaged ones.

At least for me, CF is fine; recaptcha is the only one I really have problems with.

I dont care what recaptcha wants to optimize for. I dont think that using a vpn is that a rare thing anyway. If others have figured out how to do it without requiring spending 30 seconds to solve a captcha, I dont see why websites still use recaptcha/captchas for that.

And that it is "my fault" not being logged into google I was least expecting to see here.

Imo the worst is recaptcha. At least with cloudflare the work you have to provide is minimal. With recaptcha it can take me much longer than 5 seconds, and lately I have trouble even completing their challenges correctly. Nowadays if I see a (recaptcha) captcha I drop the site unless I must visit it for some reason, it is not worth the time, the effort or the annoyance.