This was updated today, by removing old links. But that's not a good way to fix linkrot; it's better to keep the broken links, so people can look them up on the wayback machine if they wish.
> - The process to "promote" fragments of a notebook into being reusable functions seemed very high-friction: basically you're rewriting it as a normal Python package and then adding that to Jupyter's environment.
> - There aren't good boundaries between Jupyter's own Python environment, and that of your notebooks— if you have a dependency which conflicts with one of Jupyter's dependencies, then good luck.
The best Jupyter UX for me now is VSCode. Just put an .ipynb file in your workspace and you get the notebook interface inside VSCode. Put `%load_ext autoreload` and `%autoreload 2` in the first cell, and use the same python environment you're using in your workspace for the Jupyter kernel. Then you can import libraries from your project, use them, and it's very easy to promote code from the notebook into a library. You can just cut a function from the notebook, paste it into a library, add an import, and rerun the subsequent cells to verify it still works as expected.
You could do instant flood-fill on a slow PC in the early 90s. There is no need to precompute this. The possible issues are:
1. You're using a slow algorithm. This is almost certainly the case, looking at how slowly it run and at the order in which pixels get painted. The Wikipedia page on flood fill is enough to find a good algorithm.
2. Possibly, the overhead of individual canvas operations is high, so setting each pixel at once is slow. If this is the case, it would be partially ameliorated by using a span-based algorithm, but you could also run the algorithm on an offscreen byte array and then blit the result to canvas.
I would bet money that a 90s state-of-the-art algorithm running in JavaScript on an offscreen array will be perceived as instantaneous on a modern computer.
WebGPU availability factoid is actually just statistical error. Noscript Georg, who lives in a cave and browses over 10 sites a day using Emacs, is not an outlier and should have been counted.
Is the waitlist FIFO or does it prioritize? For example, suppose a Googler signed up for the waitlist, and since they already have ~all Googlers on it, they won't get access before 2026. Could they get access faster by quitting their job, deleting their Waymo account, then re-enrolling as unemployed?
It may not cast a shadow directly on it, but the bridge still destroyed the mansion's utility.
The alternative would have been to run the new east span south of the old one, instead of north. Apparently San Francisco preferred that option, but Oakland wanted the north alignment. I'm not sure why, all I've seen mentioned so far is that they chose that particular alignment to ensure drivers would get a good view of San Francisco while driving west...
> I'd say the common thread is that they expect you, the tenant, to not understand your rights and the laws that enforce those rights.
I think there's also a cost-benefit evaluation when it comes to deciding where to make use of those laws. I've heard that there are tenant screening services that will report whether you have previously sued a landlord. Is it even worth it to sue if it's going to make it more difficult to find a better place later?
It wasn't a letter of the alphabet any more than the dollar sign, the pound sign, or punctuation were letters of the alphabet. It wasn't used in words. All that happened is that someone decided to put it at the end of the alphabet song to teach it to children, and that's where it got the name.
Here's an egregious UX error in the Apple Weather app: when you look at the precipitation map, it animates the situation throughout the day, but the current time it's showing is only indicated by the cursor in a timeline at the bottom. Your location in the center in the map, where you want to look, only shows the current temperature and weather: it does not animate with the timeline and the rest of the map, and does not show the time the map is currently reflecting. So if you look at the center of the screen you have no idea what time's situation the map is showing, and if you look at the bottom you can see the time, but not the situation. You have to stop the animation, manually drag the cursor to "now", look at the storm pattern, then manually drag it an hour forward, look at the map again, etc.
Not to detract from your general point, but are you sure your example corrupts memory? AsciiStrToUpper returns a std::string (it has to), which in your example becomes a temporary object. Temporaries are destructed at the end of the containing full-expression, which in this case is the whole assignment expression. So StripAsciiWhitespace returns a view into a still-living temporary, and foo2's constructor allocates memory and makes a copy. Only then is the temporary deallocated.
Now, if you wrote absl::string_view foo2 = etc., you'd have a dangling view for sure. In practical industrial usage, you'd build with an address sanitizer (it's built into clang: -fsanitize=address), which should catch that issue.
How do you think a heavy fine works in practice? You tell the bike thief he needs to pay $1000. He doesn't pay. Then what? You garnish the wages from the regular job he doesn't have? (His actual job is bike thief.) You send bounty hunters?!
Let's say you finally manage to impound his car. Now there's an article about how people's lives are being ruined because of fines and mere misdemeanors. We're back to square one.
If you put the bike thieves in jail, sure. And when your e-bike gets stolen, you may well think you want them in jail. But the next week, the New York Times writes an article about how there are too many non-violent offenders in jail and the incarceration crisis is a national shame. Now the e-bike owner is outraged and wants the thief out of jail. It's not trivial to come up with a punishment that will deter the thieves without also deterring the voters from seeing it administered.
The grammar _is_ wrong. It should have been "We found the 'an' neuron in GPT-2". Given the article's contents, it's hard to believe that the authors would make such a mistake; it was probably done deliberately, as clickbait.
> It is the subject of much debate whether this was a trend driven by consumer preference or the fact that automakers can charge a lot more for bigger cars where the profit margins are significantly higher and therefore consistently invest billions of dollars a year in marketing campaigns to convince people they want or need bigger cars.
No mention of the CAFE standards that favor larger vehicles by imposing less stringent efficiency goals on large footprints? They were predicted [1] to cause an increase in vehicle size, and it happened.
IMHO the insistence on below-market-rate quotas is counterproductive in terms of the actual housing market (though it's probably necessary in terms of political feasibility). The BMR housing is _already built_: there is plenty of old, un-renovated housing in cities like San Francisco that ought to rent for far less than it does, were it not for the fact that quality housing is in such short supply that high-level demand spills over.
Concrete example: does your SF apartment have poor insulation? Single-pane windows? Creaking floors which are not level? In continental Europe, _that_ would be your low/medium income housing. Instead it's full of tech people because they have nowhere else to go.
Overwatch launched in 2016 to critical and commercial acclaim. I would say the opposite, it's in the last few years that Blizzard seemed to really go off the rails.
I use the Concur app to scan receipts for travel expenses. It has a useful feature called "ExpenseIt" (maybe a company they acquired?) that parses the receipt photo, finds the expense amount, and automatically files it in the current expense report. The UI is a bit jarring: you select your receipt, press the button, and after a few seconds it disappears. It doesn't take you to the expense it has created for it in the expense report, or offer such a link. But there is a fixed message saying "can't find your expenses? they were placed in the report".
...except that, every once in a while, the receipt just disappears. The receipt picture is gone, but there is no corresponding expense created in any report. As far as I could figure out, it was simply deleted.
So, whenever you scan a receipt, you have to manually track it down to the report where it placed it, to make sure it actually went somewhere and did not get deleted. Or you keep all your paper receipts and spend two hours going through them after the trip.
I tried filing a bug report but quickly realized it was impossible.