@rebane2001

Is this practical?

Not really, you can get way better performance by writing code in CSS directly rather than emulating an entire archaic CPU architecture.

Beautiful.

@blinry

It originally ran Sqeak. Sqeak is a modern Smalltalk (though Pharo is positioning itself as a replacement). It was also inspired by the DynaBook, which was another of Alan Kay's projects.

Smalltalk environments all let you inspect both the source code and the state of running objects.

For Étoilé, we built a persistent object model with some common interfaces and the UI framework exposed the same introspection APIs, so you could attach an inspector to any object and see it in a generic way, but then attach an inspector to the UI for the model object, and then to that in turn and have inspectors all the way down (or up, or something).

@danirabbit

But 'we will do the bad things, but slightly more slowly than the other team' has been a winning electoral strategy for decades!

This is what happens when people look at Duverger's law and don't go 'well Duverger clearly had zero understanding of game theory or psychology, and look, here are a load of counterexamples from the 20th century' and instead go 'yup, that's exactly how politics works in FPTP systems'.

@em_and_future_cats @rasterweb

And even if it were effective, that's a terrible argument.

I learned how to use a computer using a BBC model B. Yes, some of the things I learned there are still relevant. But I have worked with people my age who learned a decade later and have no disadvantage.

People who jumped on technology bandwagons later often have an easier time. They don't have to keep relearning things that rapidly change. While the technology is being deployed, they are solving real problems. Once the technology reaches a generally useful point with some stable features, they can jump on and adopt it at a much lower cost than the early adopters.

Look at the eCommerce boom, for example. A very small number of companies like eBay and Amazon did very well from being early adopters but a lot of the early adopters crashed and burned. In contrast, a lot of companies jumped in after the .com crash and built sustainable business models and did very well. In the UK, the supermarket that adopted online ordering earliest was Tesco (they were on Compuserve!). The two with the largest market share are Lidl and Ocado, neither of which entered that market for the first decade.

@Natasha_Jay

Real books are ones with no imaginary component. And who would want to read those?

@futurebird @meganmariehart

(I favourited and boosted this, but given the content of the post, I feel like that could be interpreted as a passive-aggressive attack. It isn't meant to be, I have nothing of value to add to this, but I completely agree with the sentiment expressed.)

@0x09 @astrid

That is the worst pun in the thread, I applaud you.

@sysop408 @jalager @kwayk42 @astrid

I used to have a flashlight that took 4.5V batteries that were actually three A cells in a bigger box. Unlike the 9V ones, which are quite hard to dismantle, these had a top that was very easy to pop off and so you could see the wiring inside.

I've never seen an A cell anywhere outside of these ones, but when I was young you could buy the 4.5V ones in most hardware stores. I haven't looked for one for a very long time, so I've no idea if that's a good thing.

@csstrowbridge @SallyStrange

I considered buying some when it was $10 but considered that the supply of greater fools was probably close to exhaustion. Apparently I underestimated the idiot supply by several orders of magnitude.

@monospace @kde

It's similar to charity work

So qualifying F/OSS projects will require a load of compliance work that doesn't make sense unless your turnover in donations is large?

I'm not familiar with what counts as volunteer work in general for tax / legal purposes in Germany.

@kde

It's hard for me to see what that should look like. If I work on a personal project with no other users, is that volunteer work? How is that different from working in my own front garden (is that volunteering in Germany)?

If my employer releases my work as F/OSS, can they count my time as volunteering for any tax purposes? What about if it's a F/OSS component that is useful only in conjunction with some non-Free thing that they sell?

@hiisikoloart @bloor

In theory, it’s people who care a lot about audio quality. They often claim to have better than average frequency range in their ears (many do, but a lot claim to hear things only bats can actually hear).

For a long time, a lot of consumer audio equipment was pretty terrible, so there were real reasons for wanting something better, I remember listening to a CD that I’d heard many times on my CD player and ripped to my iPad and discovering that CD player from the ‘80s had completely lost a load of low-volume bits and there was material that would probably have been audible on an expensive player in the ‘80s and was easily audible on a cheap player in the early 2000s.

At the same time, the Loudness War happened. Music execs found that people were more likely to like music if it was loud the first time they heard it. So they started making CDs louder. But CDs have a fixed dynamic range, so making it loader lost detail. They couldn’t do this with records because the needle would jump out of the track, so we had a weird period where LPs had better audio fidelity than CDs. Unfortunately, LPs are really finicky and it’s very easy to scratch them if you don’t perfectly balance the stylus to avoid more than minuscule pressure on the surface.

So, to listen to the highest-quality music, you needed a moderately expensive record deck, a decent amplifier (and pre-amp: again, LPs are annoying to play), and speakers. And it was fairly noticeable if you got any of these wrong.

But then DACs got a lot better. Cheap USB audio adaptors for computers had much better precision than anything available in the ‘80s, and could be placed outside of the case and away from RF interference from the computer. AAC audio supports a variable dynamic range (so bumping the loudness is just a scaling factor, not a loss of precision). Baseline speaker and amplifier quality improved a lot. By the mid 2000s, fairly cheap equipment gave better sound quality than anything you could buy in the ‘90s.

By then, an entire industry had grown up to cater to people who wanted the best sound quality possible and an even larger group of people who wanted to be seen as having the best sound quality. It moved from music appreciation to conspicuous consumption as a primary market driver. And that made it a ripe target for scams.

For analogue things, there were obvious things you could sell, like cables with gold-plated connectors. Gold is a good conductor and, unlike copper, doesn’t corrode, so this would make a difference (whether the difference is audible is another matter). But the move to mostly digital paths made this harder. You got very silly things like ‘audiophile grade’ Ethernet cables and optical connectors, which ignored the fact that the digital protocols had built-in error correction and that audio is staggeringly low bandwidth in comparison to other things carried over these connections so there’s space for a lot of error correction. A load of these things can be run over a wire coathanger with no loss in quality.

The entire ecosystem became dominated by very silly things. But they’re all quite interesting because they have some plausible-looking science behind them, which then goes off in a nonsense direction. For example, Ethernet is an electrical protocol, so signal quality matters. Gold is a good conductor. Gold connectors on Ethernet cables will reduce signal degradation. Pay no attention to the fact that the Ethernet standard is specified based on specifically rated cables and won’t be any better on ones with marginally better connectors.

My guess from the picture is that someone has noticed that electrical noise from a power supply can be a problem and has built something that looks very plausibly like it would solve that.

@arafel @cstross @bjn @lauren

Playing Russian Roulette with a six shooter would give them slightly better odds than they want to give children. Load another couple of chambers with blanks (won't kill, will burst an eardrump) and you might be getting close.

@tante

• Microsoft: How garbage can we make office and people still pay any markup?

I don't think it's fair to blame the Office team for this. I think the process in MS went something like this (having seen some of this from the inside):

Azure (which, remember, was Nadella's baby: he was head of cloud before he was CEO) noticed that cloud demand was slowing. They needed something that spurred demand.

Google built an 'AI supercomputer' out of TPUs for DeepMind. Look here's a thing that requires enormous compute requirements and even custom silicon. It's a workload perfect for the cloud!

Top-down instruction came that everyone should look at deep learning approaches for things.

OpenAI made an impressive demo and showed that they needed huge amounts of compute to train the model then more for inference. This is perfect for Azure! The start of the circular financing began, with MS investing $10B in OpenAI, which was promptly used to buy $10B of Azure compute.

Internally, investing in OpenAI was pitched as a great thing for a bunch of reasons. OpenAI was ahead of Google, so this let MS overtake Google (yay!) and claim a huge customer for Azure.

Azure's priority shifted to AI. Generic compute was out (growth there was stagnant anyway, mostly led by giving great deals to big companies switching form AWS in the hope of locking them in), AI was in. And now Nadella was going to have a bigger AI supercomputer than Pichai and wave it in his face.

AI is bringing in 'revenue' (pay no attention to the financial shenanigans) and is a big growth thing for Azure. The company's priority is now to make sure that there is demand for the thing Azure is selling.

Every business unit is told to build AI into the things. If it succeeds, great. If it looks moderately good, competitors will need to build something equivalent and that drives growth for Azure.

Everyone's bonus and promotions depend on how well their projects are adopting AI. Microsoft has long incentivised lying to management and this got much worse. Everyone who wants a bonus this year tells their CVPs that they have done a load of AI. The CVPs report to their EVPs that they've done way more AIs than all the other business units. The EVPs tell Nadella that AI is revolutionising the company.

Nadella hears a strong message form all parts of the company that AI is transformational. Any product that he touches that hasn't been sprayed with AI sends questions down the management chain about why not.

@LevZadov @catsalad @Em0nM4stodon

Come for the cats, stay for the donkeys.

@Em0nM4stodon By being easily reachable on a platform, you are actively incentivising people who wish to reach you to be on that platform.

@futurebird @Anke @va2lam @bucknam

When I was four, I learned both methods, but the second was much easier if I started with the first. The knot helped hold the thread in place for the first few stitches. When you are just learning, it’s very easy to accidentally pull the thread through when you go back, the knot stops that happening.

@neil @mansr @bojanlandekic @flux

TCP rate limits based on the delay of the acknowledgments and most consumer routers are pretty good at ensuring some modest fairness between streams.

The main failure mode is in the opposite direction with asymmetric home connections. If you’re doing a lot of uploads, the TCP window replies get delayed and that tanks the download speed.

@jwildeboer There are some companies now recycling EV batteries for home storage. For fixed locations, the efficiency Is a bit less important (space and weight are at less if a premium than in a car), but cost remains important.

@datarama @futurebird

Here, we got two separate grades for written work: One for content and one for presentation (which basically meant penmanship)

The thing is, these are not separable concerns. People who struggle to type tend to write code that’s hard to read because typing comments and using meaningful function and variable names is a struggle. So you’re implicitly measuring input-device proficiency when you judge code (this is slightly different with autocomplete because using a meaningful name ones is effort but then autocompleting it may be easier than autocompleting or typing a shorter one).

Exactly the same applies with essay writing. Writing with a pen has (for me) a much higher cognitive load. When I type, I frame a sentence in my head, edit it a bit, then flush the buffer asynchronously through my fingers without any involvement of my conscious mind. When I write with a pen, I have to think about forming the shapes. For my mother, it’s the exact opposite. We both have a thing where we can’t spell a word if you ask us, but get her to write it or me to type it and it will be correct.

If you had to type an essay using my nose, no amount of separation of presentation and content marks would save you from a low grade. For me, using a pen is different in degree but similar in kind.