bayesnet

The piracy thing is a bit backwards I think. Piracy is a consequence of subscription price so higher prices induce more piracy. The only way piracy doesn’t exist is if all content is free to begin with

> … for technical users it's a pretty noticeable step down in terms of how you may use and configure your own system

It’s such a frustrating experience on both ends—I use macOS for most of my work because I like the hardware it ships on, but it is so annoying to not be able to, like, use a different window manager (hacky accessibility API systems notwithstanding).

On the other hand I’ve probably burnt weeks troubleshooting and configuring different compositors/window managers on Arch (and STILL don’t have the setup of my dreams), so it’s frustrating on that end too.

Passport sizes are standardized and AFAIK every issuer complies: https://en.wikipedia.org/wiki/ISO/IEC_7810

I don’t understand how this was intended to work (not technically but as a “free” service). They want ISPs that don’t support v6 to pay them. Their free trial (for ISP) says

> Your free trial starts when the frontend receives the first connection from your IPv4 range.

> Registering on this page means that we can send you an email to inform you, that your free trial period is about to end. If you do not register and we are unable to find usable contact information for you, then your customers will experience service disruption trying to access sites through this frontend.

But as a website owner using this service I would just want it to work regardless of whether the users ISP is playing ball with Netiter. And there’s the coordination problem with the IPv6 transition in a nutshell I guess.

Also, “Netiter have allocated bandwidth for free trials until the end of 2014,” so I suppose this doesn’t work anymore anyway.

It’s a bit of an ominous quote given that Hilbert’s program was dismantled shortly thereafter by Gödel…

The “– Whales” part of the HN title (not in the original headline sadly) makes it seem like the whales are the source of the story. Sadly not so.

I’ve written a lot of Lean for economic modeling (so take this with the caveat that it’s not frontier-level mathematics research) but I think this problem is overstated. If you follow good engineering standards—keep primitives composable and design abstraction well—it’s not so hard to understand enough Lean to ensure the formalized statement is correct.

In part this is possible because mathlib is very well-designed and has a very good API (in no small part because they’re willing to make breaking changes all the time), so building on top of it makes life much easier.

As someone who doesn’t write Zig, I wish the code examples gave type annotations for the writers so I could see the difference between the old generic API and the new `std.Io` interface.

Yes, totally agree—I love the car and for nearly all of my driving it’s great. It’s just that the residual of long drives is more of an issue than some say because of how inconsistent the experience is. For someone my parents age, for whom things not working as expected is more befuddling and aggravating than for me, I just don’t think it’s tolerable (and they do many more road trips in retirement anyway).

I’d be happy to get the advertised charge speeds on my Equinox. The number of times I use a charging station that delivers ~120kW at most is maddening.

And it’s this kind of thing that stops mass adoption. When you fuel an ICE car, the pump just delivers fuel and that’s that. The price of fuel is prominently advertised and the experience is the same no matter the station.

Charging on the other hand is an exercise in frustration. Charging stations have different advertised power delivery rates; often they can’t meet those rates (is it the charging station? Is it your car? No way to know!); pricing is opaque; charging can stop seemingly at random .. and then you get levied the penalty congestion rate for using a charging station.

My parents asked me if they should get an EV. I advised a conventional hybrid. They are too old to have the patience to deal with all this.

I think that’s uncharitable. The study has a very specific goal of assessing whether blue light can keep seniors with problems staying awake from falling asleep early. It’s neither obvious nor intended by the authors that conclusions about this population generalize further.

Especially in the context of the flux and night shift, which are intended to help people with the opposite problem (who want to sleep sooner rather than stay awake later), it’s an important distinction.

Well, we know that conclusively in the study population of 65+ year olds who suffer from insomnia and suffer from excessive daytime sleepiness

I think it’s easier to see if you drop the time indices and just introduce a variable for the elapsed time. So:

Mary := 24 = 2(A - Δ)

A = Mary - Δ

And then it’s just substitution. But of course what appeals to one’s intuition is very personal

It’s ironic that the README has all the tells of being LLM-written: > 53 numbered rules, 9 sections, written in 1983 by people whose readers die when a sentence is ambiguous. The ones doing the heavy lifting: …

Not really a promising tell for a writing skill, IMO.

Very nice! But I was struck by this:

> The last time I had seriously dabbled with exploits and low-level recovery was when I spent months trying to revive a bricked M1 Max MacBook Pro. Several companies and repair shops had told me that getting through the problem was impossible, but after months of working at it, I finally got the machine running again on New Year’s Day 2024.

If the author is here would you please write up that story? I would also have thought it an impossible task so I’m fascinated by how you did it.

What’s the status of the OpenD fork?[0] I see Adam is still active there but I’m not sure if it’s had any impact beyond losing him as an upstream contributor.

[0]: https://opendlang.org/

I buy rice from an online store that carries speciality Japanese varieties. The rice is very good.

But this year they replaced their previously obviously-English-as-second-language-but-informative descriptions with just total slop that say absolutely nothing at all about the rice you’re buying. Every description is just platitudes about how good the rice is.

So as in TFA we get people who seemingly can’t resist “pressing the button” and getting nice shiny results even if they aren’t, like, good. It’s nuts. FWIW I emailed the rice shop and politely asked them to revert to the old descriptions but received no reply.

That auditors are responsible for catching the lies of an audited company on penalty of being suspended is a position that the big audit firms would not agree with. They might agree that they are responsible for ensuring the material accuracy of accounts if fraud occurs, but even here EY has disclaimed responsibility if the fraud were sufficiently complex [0]. Indeed auditors lobbied very hard against being mandated with a broader anti-fraud role [1].

Admittedly this is on the "real" audit side and not the advisory/consulting side, which would be the ones to handle SOC I imagine, but nonetheless a position I find a bit absurd.

[0]: https://www.ft.com/content/a9deb987-df70-4a72-bd41-47ed8942e...? [1]: https://www.ft.com/content/c25de9fb-a808-4946-ade6-80d76a66a...

Right now the HR is showing 48 bpm, so this has the fun side effect of demonstrating that the author is in pretty good shape.

The page appears to be dark themed even without dark mode enabled; if I use Noir to force-add dark mode, then the diagrams indeed become difficult to read but this seems like a Noir issue more than a site issue

Especially interesting because (as one of the compiler team members noted on Zulip) this is a miscompilation relatively easy to stumble into in safe code (yikes). Looks like the cause was a late pass in MIR optimization pipeline — I would think these are carefully vetted for soundness so am surprised that this slipped in there.

Atkinson Hyperlegible is an incredible font family! I read much faster and strain much less when I use it. The Braille Institute licenses it under SIL and includes the Glyphs 3 source files used to build it with the download too which is very cool.

I do not think that it is as well-reasoned as that. Far more likely IMO that teams are being evaluated on upsell uptake and so push these requirements to make the number on the dashboard go up.

Anyway, I’d be more sympathetic to this way of thinking if windows wasn’t getting worse over time. But it seems like investment in the OS is being disfavored for investment in AI-in-things-no-one-wants-AI-in, which is inverse of the subsidy direction you propose

I don’t think I have ever seen ∋ used to mean “such that” so I was very confused until I got to the explanation (as it were; why CONTAINS AS MEMBER is being used to mean “such that” is never explained).

To be fair I think the success of JS is in spite of it not working super well

I have a TS3+ that broke about 18 months in. I talked to support, set up a repair, and before I could send it the dock unbroke and had worked since. Truly mysterious and left me with a sense of unease with that thing given the cost.

The only surprising thing about the Tom’s Hardware example was that John Gruber evidently does not use an adblocker

> Not a product launch. Not a keynote. Not a revenue metric. The thing he was most proud of in four years at one of the biggest companies on …

What’s tiresome about this is that people don’t even bother to edit it. I use LLMs to draft long-form text all the time because I think the hardest part is getting something on the page to refine. But I would be embarrassed to leave LLM tells like this in the final result, if only because I want people to know that I actually cared about what I’m asking them to read and that I value their time.

It’s especially ironic here because this is about lauding a person who cut through the impersonal behavior of a large organization. Evidently this person was not worth even an editorial pass over the article though.

It’s your last point that keeps me up at night. I opt out of letting my data be used for training (and don’t use tools like Gemini that make this very difficult or impossible).

But I have very little reason to trust that they actually don’t use my data. The incentives to violate the data controls are so great with the billions sloshing around AI companies. The worst that would happen if they are found to be abusing my data would be civil penalties (i.e. fines) that I’m sure executives would see as the cost of doing business.

Ideally there would be an attested zero data retention option, but I’m sure that’s not available.

Really love this article. Opens with the problem statement and jumps straight into the investigation. Thanks for a very enjoyable read (and an rss feed!)