exmadscientist

They're all horrid to step on, so that's not really worth discussing.

The US plug (NEMA 1-15P is the 2-prong version, or 5-15P for the most common 3-prong grounded) is actually pretty decent. They stay in sockets pretty well if the sockets aren't bottom-of-the-barrel garbage products which, sadly, many things in the US are in general.

What many people miss when comparing US and EU plugs are that they're actually doing fairly different jobs. The 120V mains in the US is significantly less trouble than the 240V in Europe, so you can get away with a lot less mechanically on the plug. The US NEMA 6-15P plug design (the 240V version of the standard plug) is, accordingly, kind of shit. But at least it's rare enough that receptacles tend to be of decent quality, which mitigates a lot. The locking versions (L5-15R etc) are actually quite good, and should almost always be used when possible and appropriate. They're good to 480V and even a bit beyond -- not just working but actually good -- and you can't say that about most EU plug designs. You guys have those horribly bulky pin-and-sleeve disasters and we get nice NEMA locking plugs for our three-phase.

One of the other things that European plugs have to deal with, and US plugs don't, is that European electricians have no idea how to separate hot and neutral wires and basically wire them up randomly. This has major consequences in a lot of product design and is really really annoying when you're trying to build something safe. (Namely: in the US you're often (but not always) not allowed to fuse neutral wires. That makes sense, for safety reasons. You have to fuse hot... but in Europe you don't know which one is hot, so you have to fuse them both. Which means you just fused neutral. Design conflict!) That shows up in the plugs, too; with no idea where hot is, you have to be more careful and can't use neutral to help you out.

Really? I thought they were trite and not particularly engaged with the source matter. Why did you like them?

The one at the beginning was far more interesting.

> (I'm guessing)

Please don't do this. It's not helpful.

One of the most basic safeguards -- necessary, not sufficient -- for AI and even general technology in healthcare should be transparency. Keep the model a black box (if you really must), but every single input, output, and decision should be logged and available to the patient on demand, no questions asked.

That makes it a lot easier to deal with things like the bad AI transcriptions going around. Get burned by AI that can't hear what you're saying? It's in your file, what it heard and then what it screwed up. Providers don't like everything being on the record like that? Maybe they shouldn't be trusting that tool.

Want to know what the useless AI saw when it sent you to group therapy for acute mania? It's in your file. Providers don't like everything being on the record like that? Maybe they shouldn't be trusting that tool.

> bipolar

I think this is more relevant than people give it credit for. While bipolar depression and unipolar (major) depression might generally be functionally indistinguishable from the outside, they seem to have genuinely different origins within the body and brain.

And the drugs seem to be far more powerful for people with bipolar.

That's both good and bad. Good, because between mood stabilizers and very careful use of antidepressants, many people with bipolar can genuinely achieve complete treatment, or close enough to pass as complete. Which is wonderful! Bad because antidepressants and (hypo)mania are a dangerous combination, and the antidepressants can have more kick than one would normally expect. They'll often cause a patient's first manic episode, which can be very bad on its own but might get things on the right treatment course. (Of course, they often still do nothing much at all.

I think the tapering may be related as well. I was eventually diagnosed with bipolar II and I've never really had to taper off anything that didn't have a black-box warning about tapers (either intrinsically or due to blood-pressure effects, which are worth taking seriously). I've stopped some things cold which one really shouldn't stop cold. I didn't really notice much other than the bad side effects causing the stop going away.

> I do have "extra money" and am able to pay extremely qualified and expensive doctors

I don't really have "extra money" but I still do this anyway. My psychiatrist is $500 an hour, out of pocket, no insurance accepted period. He actually listens. It's worth it.

The nice thing about "ranking" stuff for yourself and not for anyone else is that some years can have three "best of the year" and some can have none. Embrace the fuzziness inherent in the system.

Because, man, some years really were just that good.

That's part of why I like the fuzziness. And why it's a really great system looking back retrospectively. I don't keep a reading journal, but if I glance over physical books in my house or reading history on my ereader, rankings of past things read get really clear really fast.

As you said it can be kind of surprising!

I'm not sure if you're just choosing not to engage fully with what I wrote or if you're trying to make some other point, but for what it's worth I do not and did not write any of my comments here with AI.

So that wasn't really a consideration for me.

Oh, that's great! I'm going to have to remember that!

> Why this word, this scene, this character, and not some other one?

The "why this not that, even when it looks like that would do just fine" style of differential analysis is an exquisitely powerful analytical tool. Asking this question is probably the single most powerful method I know of to generate deep thought, in absolutely any aspect of life. If you haven't tried it... go for it!

(Sample question: why'd I choose the word "exquisitely" in the above sentence?)

> In the short term, thinness prevails. ... The real question is not “which of these things looks nicer when you look at them for two seconds?” but “which of these things will stand the test of time?”

That tracks with another thing I've noticed I approach differently. I've recently joined a book club and everyone loves to talk about how many "stars" they gave a book. I've never really been able to rank things like that. Aside from there being no such thing as a total ordering of books (or movies or ...), who's to say what five stars from you means, compared to five stars from me? Instead I've started thinking about books in terms of "book of the month", "book of the year", "book of the decade", "book of the century", et cetera. Is this going to be the best thing I read all month or all decade? Am I going to remember this for a year or forget most of it next week? Is everyone still going to be talking about this one in 80 years? I think that's a much better scale than "one to five stars (but actually three to five stars)", and much more interesting. It also bakes in both the inherent subjectivity and inherent fuzziness really well.

(Yes, some things... plenty of things... are thin enough to be candidates for "book of the day" or "book of the hour". Few things are forgettable enough to go below that, though there have been some attempts; mostly at that level things underflow and you get in to Mystery Science Theater 3000 territory, which is its own special kind of thing.)

Exactly.

And if an appellate court got involved and somehow found a judge outsourced their judging to an AI, it's hard to imagine them accepting that. A "polite suggestion" that the relevant bar or judicial association look in to something, coming from a federal Circuit Court of Appeals, is neither polite nor a suggestion, and will receive the due gravity of a "polite suggestion" coming from one of the most powerful bodies in the country.

Are you trying to get to a working simulator, a working core in HDL/FPGA, a thoroughly debugged and validated core in HDL, a fast core in HDL, a validated fast core in HDL, or functioning silicon in hand?

I'd guess low six figures for the beginning of that list and mid eight figures for the end. Real silicon also needs all the uncore stuff, which is easy if you don't care how it performs and very very hard if you do.

Very true... but I can't help but note that basically every person I've ever met with significant experience with multiple architectures at this level, and not connected in some way to RISC-V, utterly hates the thing.

That feels like a "don't release yet" flag to me!

But I hear you.

I said "me" and "four weeks" because laying out an ISA is a pretty straightforward job for one person in one month.

The deliverable would be a (theoretically) complete specification PDF like RISC-V's.

I have little doubt that I or Adrian could do it, or plenty of others. It might not reach the quality of something like AArch64 with that level of resources, but it's not hard to beat RISC-V.

It is the easy part. Four weeks is plenty of time.

That's why it's so bizarre that the RISC-V design is so awful.

The hard part is the software support side (though, as comments elsewhere in the thread point out, AI is pretty helpful there) and then those lovely pieces like specifying the precise behavior of interrupts.

If it had a snowball's chance in hell of going in to any kind of production anywhere, I'd have no problem spending the next month laying out an ISA. But, again, as this thread makes very, very clear: ISA really just doesn't matter.

> You don't need to probe what hardware you're running on because you know being the manufacturer. The code is bespoke for your solution and nothing more. No foreign code is going to run on it.

In practice, this is not the case. The scenarios mentioned in the article involving binary blobs are pretty common, as well as other similar scenarios.

Really, I'm going to go out and say it bluntly: it is just completely freaking stupid to make an architecture where everything is optional but you have no way to query what's present. If you're going to go the optional-pieces route, you have to have a query mechanism of some sort. As the article explains, you cannot even trap instructions on RISC-V to figure out what your core supports, because bad instructions might belong to some other option. Complete. Idiocy.

> After being asked for the Nth time to explain, I decided to put it all down in one place so that I could simply link to it when asked next.

Bookmarked, because I've needed the same.

The worst part of all this is that they really should have known better by now. In 1980 you could make these kinds of mistakes, because this was pretty new territory. In 2020, doing this just makes you stupid. Or ignorant. Or both.

I haven't served in the military, but I have been onsite in awful places at someone else's schedule and whim.

Getting extended arbitrarily is horrible. The only thing worse is when they let someone else go home early, and force you to stay for no real reason. (Not even drawing straws or similar.) That genuinely broke me. Doing this to people is cruel, full stop.

(Fortunately, for as bad as it was for me, it did at least make it impossible to ignore my mental health, and was the beginning of a long, long road of treatment.)

There's massive differences between "my site isn't hugely popular, but I get some readers", "my site is up and findable but genuinely no one visits except scraping bots" and "it's available, people would visit if they knew it existed, but they're not being offered it, they're being offered bot distillations with no reference back".

The last one is the worry.

The middle is... where we all start.

The first is not a bad place to be, all things considered!

It knows you, but it doesn't know me. Perhaps you are legitimately noteworthy enough to not worry about this!

My experience being on the searching end is that these things are terrible about attribution of where they find anything. Which has bad consequences not just for authorship, but for correctness (which is the usual reason I'm poking at them -- they're being wrong again). This makes a lot of sense when you consider the massive, massive compression that's got to occur during training, but it's still frustrating.

The danger that's concerning people (rightly or wrongly) isn't that LLMs are going to be an intermediary to your website. It's that they'll be the only thing reading it. No one will ever read your post or know what you wrote. The only consumers will be LLMs, they'll train on a version that strips out you as the author (probably more due to expedience than any sort of malice; it's not like you're famous, are you?), and your idea might get embedded into a set of model weights somewhere. No human will see a byte of it.

Are you actually saying you'd be OK with that?

The problem is that (we have every reason to believe) models like Claude were trained on existing translations. Now they're being used to generate things that are direct replacements for those translations in the marketplace.

This is basically an anti-AI author or lawyer's dream case.

And, seeing this uncanny-valley-of-slop output, I kind of hope they go for it.

That shield is a drain wire combined with either a woven braid (most effective at lower frequencies) or a solid aluminized mylar foil (most effective at higher frequencies), or, for the best cables, both together.

A USB-C cable really should be shielded. Here's a cross-section I took of a nice full-featured USB-C cable: https://ibb.co/svR5Yvyr . You can see the VBUS/GND (fat red/black), the SS pairs (white plus red/blue/yellow/brown, each one foil shielded on this cable), the USB 1/2 pair (white/green), and then the CC and SBU stuff is the remaining wiring. Notice how not just the whole thing is braid+foil shielded, but the SS pairs are individually shielded; that's really important at these speeds, and what you want to see on a good cable.

Compare with two USB 2.0 cables: https://ibb.co/BHz4v1mx and https://ibb.co/GfpJmgbx ; the first one has the shield braid, as it should, but everything else is nicer on the second one. Cables!

We had good results testing Cable Matters stuff here at work, but that was for non-critical use so I didn't go too hard on the tests. (Actually, I didn't go too hard on any of it since it was mostly make-work from a manager trying to cause trouble, but still....)

I was quite impressed by GlobTek (available at Digi-Key) and have been buying those when I need something nice. Their USB-C line are some of the best I've seen, not that I'm a connoisseur.

I'm not so sure about cheap. All the cheap ones I've tried have been pretty flaky, and ended up in the junk bin. I picked up this guy for work: https://bit-trade-one.co.jp/en/adusbcim/ and it's been pretty good. That was a while ago though, and some of the cheaper ones might have caught up.

I wish there were more domestic assembly companies that catered to high-mix/short-run. The few there are are mostly terrible or unaffordable. But it's a tough business to be in.

Your competition is JLCPCB on the one end and the boutique local companies on the other. You will never match JLCPCB's pricing, but you can beat their turnaround time. Sit down and write a list of why people use each company, and what people hate about each one. Try to split the difference, see what you can provide that sits in the middle. It is not easy. Then see if you can provide that at a profit (this is the real challenge!).

It isn't a hard business to be in, technically. But it's low-margin, high-capital, very-high-competition.

No, I mean all LEDs.

The cathode marks are basically randomly assigned. This is because each company more or less designs one leadframe/package structure per size or type of LED. Sounds reasonable, right? But LED dice aren't made of silicon. They're made of weirdo high-bandgap materials, and every color is different in its own way. From physics alone, some of the colors end up with cathodes on top of the dice and some of the colors end up with anodes on top. They all get bonded in to the package the same way. So... that marking built in to the package? That's always the "top of die" wire next to it. But "top of die" might be cathode, or it might be anode. Flip a freaking coin!

If you want to see this clearly, find a datasheet that covers an entire series of LEDs. Vishay often does this, here's one: https://www.vishay.com/docs/82437/vlmo1300.pdf . They'll tell you the marking locations for each color individually. Notice how they're all completely different? Yeah.

Could the manufacturers do better? Yes. Yes, they could. We all wish they would do better. But they do not. So LEDs are the stuff of nightmares for people who assemble boards.

Through-hole parts are somewhat better (but there are certainly rogues!), but this is 2026: the volume's in surface mount.

Multipin LEDs are no better. The SMD no-lead versions are again fairly randomly marked. The PLCC-4 types are usually "cutoff corner near pin 1", but I've seen it as pin 4... or pin 2... constant vigilance is the only solution. Constant vigilance!

> If the assembly fab had access to the schematic (I know a lot of people wouldn't want to share the schematic, but if we suppose that's not a problem), would that clarify most situations?

Haha, no, not at all; that's a disaster!

No, really, the schematic usually makes it worse. CMs don't speak schematic, as a general rule. EEs don't usually draft good schematics either. And, honestly, a good schematic usually won't have this information at all. The needed information might be found in the BOM, in theory. In practice, it's probably actually also in the PLM. If your company uses a PLM system. It might not be there either.

The actual barrier here is that there's really only one guy who has all the necessary information, and that's the guy who's loading the reel into the pick-and-place machine. He's going to have to check the work no matter what. So he often ends up having to do the work instead of just check it.

This is because what orientation is required depends on what part was purchased. Not the component type; not the manufacturer; not even the specific part number. The actual dead-ass specific part purchased, and how they decided to reel it that week at the packaging subcontractor facility. This is stable over time for the major chip vendors (unless they just issued a PCN this week! hope it wasn't retroactive!); for the Shenzhen specials, don't bother. But it is not consistent among second sources for the same part, and guess what -- your purchasing department will buy the cheapest thing you allow them to buy. And those two orderable PNs that are completely identical according to the datasheet, so you listed them as fully qualified alternates, and one is way cheaper this week for some reason? Yeah, that's a reeling code difference: same part, different orientation in the tape. And it's cheap this week because some big company (probably medical...) bought the wrong one and is unloading it instead of dealing with it.

So things end up different week to week.

And only the guy loading the reels knows which reel he's grabbing. So he's got to do the work.

> Is 100% automation possible at all?

No. No one genuinely cares. This is a throughput-optimized industry: the first board takes a while to get set up to build. Then there's thousands and thousands of them.

At the end of the day, the manufacturing data packages have (should have) centroid, orientation, and pin 1 data. That's enough to get the job built. Any more information is very likely to overconstrain things, leading to conflicting information, leading to build stops (conflicting information should always stop the build, always), leading to the engineer who did the overconstraining getting called in to fix it. Probably by not realizing what they did and saying "uh, just, uh, you know what to do, so just do that I guess".

The good professionals learn to make the manufacturing data files say that from the beginning.

All LEDs are like this. It's horrible. The only good thing about it is that every professional in the entire industry has come to expect that if it's an LED, something somewhere will be wrong with it.

Why is this posted here? It's old, and it's not terribly relevant, though it's not useless.

For anything larger than 2 pins, pin 1 at upper left and pin 1 at lower left are both quite common. Components are also reeled more or less randomly. (There are some patterns, but you'll be better served by pretending there aren't.)

Put that all together and it means... it's all a cluster and gets manual intervention for every build. You can't automate it, but you can automate the intervention (that's what JLCPCB has done to enable scaling to lower volumes).

Just pick "lower left" or "upper left" for your library and stick to it. That's the best you can realistically do.

Also, LEDs are terrible and everyone hates them.