actuallyalys

Given the severity of long Covid in many cases, I doubt that’s a more than a negligible factor.

The headline is a bit misleading, as the 1 million figure refers to children who had ever had Long Covid, and 293,000 were experiencing it at the time of the survey. That doesn't mean it isn't a serious issue, but I believe accuracy is important here.

The log suggests it doesn’t immediately stop upon noticing the file is over 1 TB—the second entry comes two minutes later when it’s trying to upload the 100,001st block. This makes me wonder whether 100,000 blocks is an intentional limit on file size or whether it’s some sort of internal threshold that isn’t meant to prevent uploading files of that size but is buggy. Perhaps files bigger than 1 TB have slightly different behavior by design and that behavior was broken?

Since 2 minutes isn’t long enough to upload 1 TB, it’s either looking at the blocks out of order and skipping ahead to block 100,001 for some reason or noticing that’s the first block that hasn’t been uploaded yet.

Another comment suggests 10 MB isn’t a fixed limit. In that case, I wonder if 100,000 blocks is the limit and the intention of the designers is that users with such large files would increase their block size. If so, that should be documented of course, and a scenario. Although it’s still a bit strange that it would apparently try to upload the 100,001st block and not immediately warn the user that the file size is incompatible with their block size.

The fact that many are transgender seems to be relevant because it’s a recruiting and manipulation tactic, not because of a connection to “trans activism.” I haven’t seen any evidence of that connection besides people involved being transgender.

“Bubble” is probably more apt.

There’s some truth in that, but isn’t making a radically cheaper version also a new idea that deepseek didn’t know whether it would work? I mean, there was already research into distillation, but there was already research into some of (most of?) OpenAI’s ideas.

Using RTL doesn’t necessarily result in large binaries. In my (non-WASM) experience, they’re more in line with the “hundreds of kilobytes” flohofwoe suggests.

Pascal can definitely produce small binaries (and it looks like it does dead code elimination, under the name “smartlinking” [0]), so it may be a limitation of the WASM backend like you said or not enabled for some reason.

[0]: https://www.freepascal.org/docs-html/user/userse13.html

They list a service for each address, so maybe you could block all the non-Route 53 IP addresses. Although that assumes they aren’t using the Route 53 IPs or unlisted IPs for scraping (the page warns it’s not a comprehensive list).

Regardless, it sucks that you have to deal with this. The fact that you’re a customer makes it all the more absurd.

Someone could say something similar about the large number of people who apparently reviewed this essay, who were supposed to critique it in order for him to make it stronger. It's possible he just ignored their criticism, but it's also possible they already agree with Graham and didn't think about the flawed premises, so their feedback didn't address what might be "blatantly obvious" to you or I.

Similarly, Graham almost certainly already has strong opinions on the basic premises of this essay. Thus, the process of revising and polishing his essay to make it readable and compelling doesn't help him spot any of these obvious critiques. As you quoted, he believes the people advocating "people of color" over other terms have no principles. Thus he can't apply their principles to his own essay and anticipate their criticisms. Based on how he describes "wokeness," he seems to think are generally unprincipled.

Neither he nor his reviewers are equipped to analyze the substance, which is why it can be stylistically strong but substantially weak.

I don't think being written in VB6 is actually a good reason not to use XYPlorer, which seems like a capable tool, but this page doesn't seem really reassuring? What would reassure me is knowing that there is a maintained version of VB6 for modern systems. Luckily, there apparently is such an implementation, twinBasic, and they are already using it for 64-bit releases.

It depends. If you need basic word processing, spreadsheets, presentations, and email, you can definitely get by with alternatives. If you want more advanced functionality, you’re more likely to run into limitations or just things that work differently than what you might be used to. Microsoft 365 is not better in every way, but it’s a strong contender.

I suspect the familiarity and compatibility probably cinches it for a lot of people. Honestly, the convenience and familiarity are valuable, even if you and I would prefer open source options were more popular.

Thanks for the insight! That makes me even more doubtful of the 30 percent claim.

In my experience, while there are some things in the backlogs that a developer can basically pick up and run with, others require input from other roles to actually address or implement properly. And then once developed, you need to test, document, and localize a feature. (Realistically, you probably want to market it as well.) Depending on your design and support model, your customer service teams may need to help organizations implement the feature effectively or may need to be trained on how to answer customer questions about that feature.

I don’t find the claim of 30 percent improvement plausible, but it actually does track to me that you might pause hiring for a time if your employees were suddenly more productive in order to catch up in other parts of the business or figure out what markets to put your new productivity toward.

> The productivity gains of 30% are probably overstated

I am doubtful as well.

I could imagine 30 percent among certain engineers for certain tasks, especially if you use a popular language with popular libraries and frameworks that are well-represented in the training dataset. I don’t know how typical of a codebase Salesforce has. They could also finetune a model on their own codebase or devote a small team of engineers to figuring out which prompts, models, etc. work best for their codebase and process. In theory, those advantages could boost it beyond what testing would typically show.

But a consistent increase of “more than” 30 percent across the whole engineering workforce seems less plausible, especially lacking details on how they measured that and uptake numbers. Edited to add: Are they even confident that their engineers are using it consistently? At this scale, that’s not a given.

I’d be interested to know whether Salesforce customers have noticed a change in the number or scope of features being announced. A change of this size seems like it should be noticeable from the outside. I’d like to hear from the engineers in particular.

Any sharing of information or data could be construed as a collaboration, but this seems to go beyond that, as the research paper includes multiple Waymo employees as coauthors.

It’s with noting this research is a collaboration between Swiss Re and Waymo. That doesn’t invalidate the research necessarily, but warrants more skepticism than if it were an independent research by Swiss Re, like the article implies.

> Although I sincerely worry that empathizing with people like this is a trap.

I have a similar sense. However, I’m not sure empathizing with them is itself the trap. You don’t want your empathy to make you overly credulous. For example, while you might empathize with him feeling directionless at the acquiring company, I don’t think you should therefore conclude that he is right about the new coworkers being “NPCs.”

Thanks for your kind words! I do think there are valid criticisms in the article, even if the overall framing isn’t productive.

While the tone and framing may make it especially good at attracting responses, I do want to clarify that I don’t think it was deliberately optimized for the engagement.

I wouldn’t suggest duplicating the id field, but moving from the data field.

> Why do you expect SQLite indexes on expressions to be worse than what NoSQL databases do for indexing?

I don’t have a concrete reason. It’s just that indexes on expressions are not intended as the main index for a SQLite database to my knowledge. Having thought about it more and read the page on expression indexes more thoroughly[0], I think it’s probably unlikely to be a noticeable downside.

The only remaining downside is that the item being indexed might be outright missing, but it sounds like that’s basically a feature in your case, as you’re opting for flexibility.

[0]: https://sqlite.org/expridx.html

It's currently just as easy to use for learning and experimenting on Windows as you describe. I don't use Python on Windows heavily enough to evaluate it for more serious development.

There’s truth in broad strokes (e.g., the lack of a stable ABI is a genuine problem) that I wish people would acknowledge. I think your attitude is a healthy one! However, I don’t think this article is written very charitably. It seems unfair to Linux developers to suggest that they spend “very little to no time” testing their changes for regressions, for example. And while AppImage, snaps, and flatpacks are imperfect, I’m not sure they deserve to be dismissed as just “virtual lightweight machines.” (I’m not sure how accurate that even is for AppImage.)

Linux users can be defensive or even toxic about their operating system so I hardly blame the author for being impatient or frustrated. Still, I don’t blame people for being frustrated in turn at that tone.

I’m curious what the reasoning behind having no id or key column. Do you use indexed expressions to index JSON fields? Do you use Rowid for certain queries? Or do you not bother with indexes?

My understanding is that NoSQL databases still have indexes and it seems like using SQLite as you demonstrate could be worse in that regard.

I’m not sure any one factor “explains” why something is written in a particular way. History, practical concerns, conveying the tone and register, signaling, and official requirements all play a part.

Still, it wouldn’t shock me if legal language is an unusual in how much of its incomprehensibility is explained by the “magic spell” hypothesis.

Looking at the full text of the article, much of the analysis involves center embedding, which makes me wonder whether other features that contribute to complexity work the same way.

Full text: https://www.researchgate.net/profile/Eric-Martinez-6/publica...

Anecdotally, it's common for elderly people to be doing relatively well and then decline quickly, so this result seems pretty intuitive. I would be curious if this also holds for diseases that aren't as linked to aging.

She also does mention things she did like about the thing she’s criticizing. For example, she liked the themed food at the Star Wars hotel, as I recall.

I’m not sure. One point I agree with the author is that it’s a complicated situation likely involving both social and technical factors.

The confusing part of this article is the argument that the kernel maintainers are strongly against RAII. While I guess it’s possible Asahi Lina is unaware of this or is stubbornly including it anyway because she believes it’s such a large benefit, it seems more likely that it isn’t actually the dealbreaker.

In fairness, I’m hardly an expert on C++, Rust, or kernel development, so it’s possible that I’m missing something.

I personally like Fennel a lot, but it’s not at the point where I would build something large on top of it. The language itself is solid, but like you said, there’s not a lot of tooling.