2,552 karma · joined November 8, 2013
On HN the main goal of upranking is if one comment is more interesting/informative than another (or as a group test to see how robust its argument is if the voter has no experience to judge directly). Downranking ime isn't meant to be the method used shift the order of comments but rather to discourage a post that doesn't fit the HN culture/guidelines. OTOH most popular gamified discussion systems don't discourage use of reactionary downvoting, which can creep into other posting cultures.
The problem the grandparent post raises is if signals that voting users would ordinarily use to shape the continued posting culture (eg: downranking comments that don't fit in tone/substance) aren't used like they were intended to be and if the guidelines discourage meta discussion then there isn't any other avenue to inform users what the desired culture should be in practice.
Certainly one can post non-meta comments showing what type of comments one would like (and thankfully for most strictly tech topics here it's still reasonable) but if the culture shifts enough among the silent voting users then the concern is this erodes the quality of discussion as the signals for what is wanted/not wanted get skewed.
Afaik it's only any Via support per se that the keyboard firmware has allowing for such on the fly customization, rather than QMK. With QMK it's all manual flashing (I was somewhat surprised that it was easier than I was expecting to customize the layout/layers/lights with code though, largely thanks to the supportive community).
One of many everything-and-the-kitchen-sink features Opera Presto had during its heyday. Others included a separate Bittorrent client, desktop widgets that could be moved outside of the browser window, full IRC client, email client and peerless hotkey actions customization.
By some miracle the browser still managed to be a rather lean binary.
Evidenced by most comments on HN at the time I posted not being about the software support per se being withdrawn but jumping off the premise it was only programmable by the US, with comment chains ranging from US software and supplied aircraft can't be trusted to going so far as talking about kill switches. This led to a very noisy, knee-jerky discussion based on an incorrect basis.
Quotes from the OP's article:
> Ukrainian pilots are not receiving updates, and the programs could soon become obsolete [...] France’s Mirage 2000s are equipped with their own powerful jammers, and the Americans are not involved in their programming.
> In the longer term, the Ukrainians could refit their F-16s with non-American electronic countermeasures [...] But that could take time and money...
Now contrast to the original Forbes citation which explained the US team doing the reprogramming wasn't even familiar with the system until being tasked to reprogram it for use in Ukraine, doing so in just two weeks and it being a multi-national effort with two European countries and an overseas facility.
Had the discourse began with that there would still be criticism of the government's decision, given the US was leading the reprogramming effort, but with clarity on what was occurring.
In the very Forbes article the OP's article cites it links to info about this F-16 reprogramming effort[1], showing it was collaboration between the US/Norway/Denmark and that the US electronic warfare team wasn't familiar with the system, yet within two weeks they say they managed to reprogram them to meet the initial deadline.
> The 68th EWS assembled a dedicated team comprised of a mixture of seasoned experts and bright, young engineers to approach the reprograming challenge. Their first task was to understand the unfamiliar EW system and how to reprogram it.
> Relying on data provided by Denmark and Norway, then adapting new processes and approaches to the usual process, the team was able to understand the system and start their work.
> After understanding the system, the 68th EWS deviated again from normal methods and sent its members overseas to a partner-nation lab to collaboratively develop and test the system alongside coalition teammates.
[1] https://www.dvidshub.net/news/479401/dominate-spectrum-350th...
Manifest v3 and Web Integrity API are prominent examples of Google's team shaping how all Chromium based browsers will be, regardless of pushback (though they relented with the latter for now).
So despite the messaging from LastPass about improving iteration count for existing users it wasn't always accurate.
[1] https://news.ycombinator.com/item?id=34152779
[2] https://www.reddit.com/r/Lastpass/comments/zuve7t/cracking_e...
The other aspect would be if Google were able to cross-reference address book contacts that were obtained from permissions from their apps/services to check what name is most commonly associated with that number.
Tbh I don't think the worst problem here is with stating something is overcomplicated but the directness and variously pointed use of 'you'. People will naturally be more defensive.
Many of the example statements involve this use of 'you' in the same fashion. 'Did you even read my email?', 'If you would have read the whole email you’d know the answer to this'. It's a deliberate use of 'you' like an intensifier.
With minor rephrasing these could just be expressed more passively, for scenarios where one doesn't want an almost completely euphemistic tone.
Though it's easy enough to run a caddy local server. It was easier in the past, in earlier caddy versions one could just place the caddy binary in PATH and launch `caddy` from any directory to auto serve a local server, now it requires more arguments or a config file.
I'd presumed it was just the VM they're heuristically detecting but sounds like some are experiencing issues on Linux in general.
Presumably this would also preserve the state of animations, such as GIFs and objects with embedded animated documents (like SVG). With current methods these get reset (along with the already mentioned video state) which break the illusion of an element being transported elsewhere when the original can't be just literally moved.
This has caused a lot of conflicting anecdotes since those finding their prompts aren't censored are running the distilled/fine-tuned models not the foundational base model.
A sibling comment was facetiously pointing out that the cost of running the 'real' R1 model being discussed locally is out of the price range of most, however someone in this thread actually has run it locally and their findings match those of the article[1].
I'm assuming you're using the Llama distilled model, which doesn't have the censorship since the reasoning is transferred but not the safety training[1], however the main R1 model is censored but since it's too demanding for most to self host there are a lot of comments about how their locally hosted version isn't since they're using the distilled model.
It's this primary R1 model that appears to have been used for the article's analysis.
Someone in that thread says Youtube confirmed they were doing this but when finding a source[1] statement Google don't directly acknowledge this but use language about having launched an effort to urge viewers to allow ads or buy Premium, whatever they may mean by that.
[1] https://www.androidauthority.com/youtube-long-unskippable-ad...
> You're running Llama-distilled R1 locally. Distillation transfers the reasoning process, but not the "safety" post-training. So you see the answer mostly from Llama itself. R1 refuses to answer this question without any system prompt (official API or locally).
It was further complicated because a couple of the law enforcement officers involved with setting up one of the six murder-for-hire scams* stole the Bitcoin Ulbricht paid and it was also felt that trying to prosecute based solely on the other chat logs would have been difficult. The FBI agent who arrested Ulbricht was interviewed about it recently[1].
* The other five are said to not have been law enforcement, which makes it curious the number of times Ulbricht was scammed in this manner.
The optimal approach is applying the appropriate `prefers-color-scheme` using CSS alone, while additionally allowing a theme override using JS/storage. Fewer do this though, even though it wouldn't require any cookies and thus no consent dialog.
The worst are sites that only have theme switching gated behind registration.
Ironically they end by saying they'll use an invert filter as a workaround, while for those who prefer dark mode if doing the same on the author's own stark white bg/black text site (such as Vivaldi's 'Invert Mode') will produce the very contrast ratio in dark they're complaining about :p