bux93

IBM CUA (1987) should be required reading https://en.wikipedia.org/wiki/IBM_Common_User_Access

For most people, it is part of the OS though. Windows comes with Client and Server services activated by default even on the Home edition for some reason. The functionality is dumbed down and hidden though, while onedrive - which requires registering a windows account - is shoved in the user's face.

Well, the dB scale was once created on the basis that 1dB was thought to be the "just noticeable difference". Off the top of my head, I think trained listeners can actually notice differences of 0.3 dB, though this sort of thing is going to be frequency dependent too.

The eq-adjustments you'll find online often have adjustments ranging from 1 to 6dB in different frequencies. That's enough to notice.

Comparing settings/devices, it's very easy to notice. Just play some music on your laptop/phone speakers and move the device around a bit, and you'll hear striking differences in highs and lows.

However in isolation, I think most people wouldn't be able to say if a particular sound source is "good" or "bad". It takes a while for you to clock that, no, it's not the teams/zoom call that has bad quality, it's your headset that's dropping mids.

djbdns/tinydns can schedule a change and give out a lower and lower TTL the nearer the changeover time gets - a quick search doesn't show any newer implementations but this is surely a good way to do it.

It's to create an in-group, and you are in the out-group.

My car has wired carplay/android auto, but I have a nifty little USB dongle that receives carplay/android auto wirelessly and passes it to the car via USB. Sometimes called an AI box, I think.

These things contain a whole System-on-a-Chip board. I presume what it's acting like a middle box; pretending to be the headunit to the phone, and then sending the phone's output to the headunit, pretending to be a phone.

Since they need a quite beefy CPU to do that, I'm guessing they don't just pass along packets, but actually speak the protocol on both ends, and perhaps transcode the a/v stream.

That's in fact one of the pros of carplay/android auto as opposed to just putting your phone in one of those windscreen attached holders; you get to use the volume buttons on the steering wheel. Annoyingly my car doesn't have play/pause buttons on there, but if it had, they would work.

Agreed. Their privacy page even says they'll remove data if you withdraw consent, but they don't ask for consent. They also don't mention any you could object to data processing, claiming that "Processing is necessary to perform a contract with the data subject and to take steps toward the conclusion of a business relationship." which is a very contorted interpretation; taking steps towards the conclusion is about making quotes and such. It makes me sour on their claims "Keep your data private, compliant, and fully in the EU. As a privacy-first European company, we help you stay aligned with GDPR. No surprises. Full transparency."

Linux is only used as a kernel temporarily until GNU is finished.

This comparison to qemu gives some idea: https://www.unicorn-engine.org/docs/beyond_qemu.html

The ability to execute and inspect some code without any context (no OS, not even a complete binary) is useful for reverse/security engineering.

The parable-replication crisis is real though

If you book Ryanair, and your concentration slips up for a second, you WILL make a mistake and it WILL cost you money and/or inconvenience you. Then you might complain about that, and be met with smarmy, smug, smarter-than-you people who insist that they fly with Ryanair all the time and never pay too much, people who can't read shouldn't be allowed to book flights, etc. etc.

It's worth avoiding Ryanair just to avoid that scenario.

I have fond memories of the, I think 1995, Compton's Interactive Encyclopedia. I think it was given away with new computers, like Encarta and Grolier's. I think I did buy the Compton's CD, but at a huge discount, since I bought it grey market (not bundled with a PC, but from a PC retailer).

Flexi discs were cool inserts for some home computer magazines; you'd dub the flexi disc to music cassette, and the noise and beeps were computer programs for your home computer.

Do you see "personal" and something that looks like your name in the path? That's your personal onedrive, like your home directory on a unix system.

See "sites" in the URL? That's a sharepoint site (AKA teams "shared" folder).

The former disappears (after a year) when the user license is removed. The latter is not associated with an individual user, so even if everyone in a team leaves the company it isn't just automatically removed.

Following was wrong and had been edited: The non-business personal onedrive was a box.com/dropbox/g-drive competitor. Microsoft moved its backend to Sharepoint at some time. (Onedrive for business used Sharepoint from the get-go). The integration of the personal drive, even though it's a descendent from the 'for business' product, is still quite unintuitive in my opinion.

Not great for t-shirts

That he backed PQ crypto that turned out to be broken later should be an argument in favor of hybrid (belts-and-suspender) schemes rather than against it. Embarassing behavior amounts to not much more than ad hominems. Hybrid KEMs are a good idea.

Only if you ignore almost every input and output that neurons have.

https://www.quantamagazine.org/ai-is-nothing-like-a-brain-an... https://pmc.ncbi.nlm.nih.gov/articles/PMC9665914/

This is why making more neuromorphic NNs is still an active area of research, although they typically all focus on another extremely simplified model (spiking neural networks).

Do toilet breaks count towards the 30 minutes?

In "peeking under the hood", it says quite honestly "SLAX is purely syntactic sugar."

I think that the XML syntax of XSLT itself was only one barrier for people to adopt XSLT, even in the XML heydays.

The main obstacles appear to me that the execution model is hard for people to graps; you need to think both as a parser (apply-templates/) and in a more declarative style at the same time. The XSLT can be in a completely different order to the document and in fact it can visit nodes in the document multiple times, hopping through nodes in a different order each time, with lovely constructs like apply-template with "mode" and "select" mixed with call-template by "name", plus you get to use xpath and for-each to boot. The control flow changes from the order in the input, to some predefined order, and back depending on when you decide to match(-template) or for-each. There's a lot going on at the same time! Fun times!

Wero is super confusing. They're in the business of acquiring different methods (I don't even know if they always buy them outright or if they merge or they are just associated in some way), branding them ALL wero, and announcing that every payment in every channel will be rolled out SOON via wero, without ever offering specifics.

So in The Netherlands wero is the new name of eCommerce payments, but in another country the new name for peer2peer. But no idea when p2p will launch in the Netherlands or when eCommerce will launch elsewhere. And if the existing services will be degraded when they are internationalized or merged.

The merchant typically uses stripe or adyen or whatever (mollie has a cute name!), a payment service provider or PSP.

The PSP looks at what methods the merchant wants to accept, which methods the user could potentially be using (based on e.g. country by geo IP or some delivery location) and show the relevant icons.

EU users will see schemes like wero or Przelewy24, Japanese customers will see 'konbini' among the icons, and US users may only see credit cards, Apple Pay and Affirm. There are TONS of payment services. Stripe lists 123 of them.

The merchant will want to exclude methods that have high costs (for themselves), maybe they also care about their customers not getting into debt (so no buy-now-pay-later or credit), and some payment methods have higher rates of disputes/chargebacks (e.g. Amex).

In general, most merchants will want to offer as many methods as possible to prevent consumers who have a preference (this week) for using account A over account B from bouncing.

I'm seeing more of a gradual path.

Chrome 124 (april 24) introduced hybrid post-quantum TLS, and Chrome 131 (nov 24) switched to a hybrid using ML-KEM, which was standardized in 2024, just after Firefox 132 (oktober 2024), while openssh introduced a hybrid scheme in release 9.0 (April 2022) and made ML-KEM+25519 default in OpenSSH 10.0 (April 2025).

Hybrid PQ schemes being adopted in other places is people playing catch-up, not the avant garde.

I'd say digital signatures should be the foremost concern, those may need to provide non-repudiation for decades.

The article states

>By feeding legacy PRG (circa 1985) and logics to models like Claude, ChatGPT, developers can now instruct the AI to translate decades-old dBase PRG directly into memory-safe Rust, highly concurrent Go, or modern Dart/Flutter cross-platform applications.

And it alludes to this early on, but it doesn't show any examples.

Or maybe it should be mandatory for all companies to pay ransomware attackers. Think of it as an involuntary bounty program. Now they get to just say 'sorry (for your hurt feelings)' and suffer no consequences.

Apart from the 4% of the total worldwide annual turnover fine that theoretically could be levied under GDPR, but has never been imposed in full.

A PhD-level toddler, mind you.

Because both of them are optimized for hardware. Neural networks, despite the name, have very little similarity to biologics.

There's a lot of multiplication of numbers in parallel, so it makes sense to try to fit that to matrices.

Cryptography is built bottom-up, but likewise it makes sense to exploit data structures that already exist in silicon.

Two words. Volkswagen sausage.

Also, this is not how ER doctors work? They are not trained for this, nor does it reflect their day-to-day performance. If they would work like this, perhaps they would know a bit more about the nurse writing down those notes, and the kinds of things that particular nurse is likely to miss or overemphasize - just as an example.

The article gives a neat example: In one case in the Harvard study, a patient presented with a blood clot to the lungs and worsening symptoms. Human doctors thought the anti-coagulants were failing, but the AI noticed something the humans did not: the patient’s history of lupus meant this might be causing the inflammation of the lungs. The AI was proved correct.

Which is nice and all, but in the presence of a blood clot, I can understand that treating inflammation instead is not the first thing on a doctor's mind, what with blood clots being potentially life threatening and all. It raises the question; was this a real-life case, and what happened to that patient? Since this is a case for which the correct diagnosis is known, it was eventually correctly diagnosed - presumably then the patient did not die of a blood clot, nor of an uncontrollable fever.

Also, how representative is a patient with Lupus? According to House, MD, it's never Lupus.

No, typically it is not.

https://en.wikipedia.org/wiki/Consent_or_pay