GLP-1s and their successors for type 2 diabetes, cardiovascular disease, and addiction (to counter Western disease from diet) and vaccines and immunotherapy for cancers (most cancers appear to be caused by viruses).
It's really very outsated in terms of modern browser features and options. It's really an incredibly minimalist version of Firefox that looks like it could have come out of the Netscape era. My main gripe is lack of DoH support.
Having said that, I keep it installed when I need a webpage to work that will load every It's meant to and will filter out nothing. It works, but you'll quickly be missing a lot of the more modern features & conviences quickly.
They have also had an ongoing problem with Cloudflare hosted pages outright blocking PM. I haven't kept up with how that's going.
Oof, that's too bad. The libraries near me are great for my toddler. They do story time and play time, and it's a good chance for my kid to play with other kids. My kiddo always checks out a book (or three) when we visit.
The sketchfab examples are fantastic, to be able to move around in a 3D space, like it's some kind of scifi simulation.
The mouse controls are confusing the heck out of me. It shows a 'grab' icon but nothing about it grabs as the movement direction is the opposite, feels completely unnatural.
This HN post is seven days old but displayed and ranked as if it was posted 9 hours ago, with all timestamps falsified. The déjà vu effect is disconcerting and an absolute mind fuck. Please stop doing this, ffs. The person who thought this would be a good idea is a madman.
> So I have been using Cursor a lot more in a vibe code way lately and I have been coming across what a lot of people report: sometimes the model will rewrite perfectly working code that I didn't ask it to touch and break it.
I don't find this particularly problematic because I can quickly see the unnecessary changes in git and revert them.
Like, I guess it would be nice if I didn't have to do that, but compared to the value I'm getting it's not a big deal.
How many will follow through with these announcements? During Trump's first term, announcing huge projects in the US and then not following through was a common tactic for companies dealing with Trump. Foxconn, for example, announced a new $10 billion factory in Wisconsin. They made some initial investments and stopped when people stopped paying attention. Instead of the promised 13.000, they now employ about 1.000 people there.
And what about all the companies that will have gone out of business by then? This mainly affects small companies, which are exactly the companies you need for a healthy economy. In some cases, they have shipments already paid for that they can't accept because they don't have the liquid assets to pay the unexpected tariffs, so these companies are now at risk of going out of business completely unnecessarily.
It never makes sense to use tariffs for economic reasons. It just does not work. Tariffs can make sense for strategic reasons if you're willing to take an economic hit to lower dependence on other countries for critical industries or technologies. However, the idea that taxes are ever "a good disruption" for the economy does not bear out.
> I'm not so convinced that I'd prefer to live isolated and "just" get a robotaxi for every excursion anyone wants to do. I'd rather my kids walked about a quarter mile to school with several neighbors. Exercise and being outside are good!
I believe that robotaxis will enable totally new behaviors. For example, if you don't live immediately near a park, you won't often go there. It's just too tiresome to use public transit to visit a park just for a short walk/run/play. And personal cars are not available for children.
With robocars, you'll be able to text your friend: "hi, meet you at the park corner in 10", jump into a car, and arrive there. This will have zero friction, so it's far more likely to become a habitual behavior.
> Even assuming we turned smarter and built clean nuclear plants everywhere, just all the paving of roads, tires etc. takes a lot of resources.
Ha. One line of Manhattan subway now costs as much as 1500 miles of modern 6-lane freeway. Urban construction is EXPENSIVE.
To be clear, the GDPR never uses the term Personally Identifying Information. It uses PD or Personal Data: this can be identifying on its own, but it’s more likely that some aggregate of multiple pieces of PD become identifying only when taken together.
I haven't run into such issues intermingling C11 features in a .m file. Possibly you just needed to set the right compiler flags, or it was a bug in older versions of clang.
>> Our first contribution is Optimised Attention, which performs similarly to standard attention, but has 3/4 as many parameters and one matrix multiplication fewer per head. Next, we introduce Efficient Attention, which performs on par with standard attention with only 1/2 as many parameters as many parameters and two matrix multiplications fewer per head and is up to twice as fast as standard attention. Lastly, we introduce Super Attention, which surpasses standard attention by a significant margin in both vision and natural language processing tasks while having fewer parameters and matrix multiplications.
>>> Convolution is in fact multiplication in Fourier space (this is the convolution theorem [1]) which says that Fourier transforms convert convolutions to products.
As someone who co-incidentally started dabbling in Astrophotography as a hobby in early 2019 before Starlink launched, back then you literally could capture a single 20-second exposure (on a very wide lens, so no obvious star trail/blur at that focal length due to the Earth's rotation), and get images with no satellites.
Now (and even in 2021 it was getting hard to do that) it's impossible to do that, even with 10 second exposures.
What's needed now is multiple exposures, and merging/integrating them in something like Siril (https://siril.org/) to remove the obvious satellite trails.
However, arguably, integrating multiple exposures, while annoying and time-consuming workflow-wise (i.e. can't just look at images directly from camera, and currently need to convert to TIFF first) is often the better way to get slightly-less-noisy images anyway, and integrate effectively longer exposures without star-trails, so it's a tricky one.
