Before Silicon Valley, a lot of tech was done in the East. It's why some of the biggest "tech" companies were based on the east - Kodak, for example. But other ones like IBM were based on the East Coast. And many early computer companies were started in the East.

The westward migration created Silicon Valley as companies like Fairchild and such created companies in the orange groves of California.

Well, I'm pretty sure your bank has a web interface, so you don't have to use the banking app to do your banking.

Your work auth app may be more complex - if it's TOTP, you might be able to export the seed and key and there are plenty TOTP generators. (It's also one of the easiest programming assignments to do so any competent programmer should be able to come up with something in under an hour without vibe coding). Unfortunately there are a few TOTP systems that make it extremely difficult to get the seed and key values to use a third party TOTP generator.

Your state legal filing app, well, I'm sure there's a web based version of that as well - I can't imagine the state restricting legal filings to owners of Android or Apple phones.

Now, the car sharing app might be more complex - I think most car sharing sites have regular websites so you can book them from the desk, but there is also a good possibility a lot of them are app only for no good reason. Though many do use app based "car keys" which use some NFC thing to control the car.

If the C code does something unsafe, how can it be machine-translated into safe Rust, without knowing the intention of the programmer?

The key word being "if".

There's a ton of C code out there, the vast majority of which doesn't do anything unsafe. Take say, Apache httpd - what unsafe thing would it do? It just forks, spits a file out over a socket and that's it. The most unsafe thing it does would be to run a plugin, but that's mostly handled by the dynamic loader.

In fact, most userspace code out there isn't really doing anything irrational or unsafe - they're using kernel interfaces to do the unsafe bits, and everything else is just system calls and library functions.

And even if you can't convert all code, you can protect most code. Never let perfect get in the way of good enough. The few bits of C code that need to exist because it's unsafe is much easier than millions of lines of C code that doesn't do unsafe stuff but can become unsafe in certain weird situations.

It's not just compensation. Compensation for electricians is quite high already - the problem is the ramp up to those levels is long tedious and hard.

Pay for apprentices is abysmally low - honestly, it's not much above McDonalds, and the physical strain and effort of the work is similar, if not more. (Which is why "join a trade" is not quite a solution - most trades are physically demanding jobs and unless you start in your 20s, you won't be able to do it in your 40s. But by then you should have enough to either be leading a bunch of apprentices or running your own business).

Also, the historic demand for transformers, breakers, and other such equipment has usually been sufficient to keep the few companies in business. If there's worldwide demand for say, 12 mega transformers a year, and everyone has surge capacity to produce say, 15 transformers a year, you don't invest in a new factory because demand is unlikely to change. Things like AI datacenters and such have driven the demand up - but it takes years to bring on new factories and they'd want a solid order book confirmed before they invest millions to make 15 transformers a year because if demand drops back down to 12, you're screwed because now you have a factory that isn't being used. And this stuff is custom manufactured so you can't even make a spare.

It's why Airbus and Boeing have 5-7+ years of orders for new aircraft (and likely closer to 10 for both). Ditto things like fabs - ASML is the only company to make it not because they're the only ones that can do it, but because they fill worldwide demand. A competitor starting up will need to make a case for competing in a segment where current production satisfies demand and future demand will not only increase, but stay increased to keep both companies viable because if demand suddenly drops, then it can be hard to get a return.

Right now the lead time on many parts is in the 4-5 year range, and it's hung around that for the past 4-5 years now showing production and demand is basically steady. Someone might be able to produce more but then orders are going to run out and now you've got no customers because you've emptied the queue.

You're not understanding the incentive structure that led to this state of affairs in the first place. To use USB would mean that they would have to do a ton of driver work. They would have to implement these generic control surfaces and they would have to ensure that their hardware does what the USB specs say it should. You'd end up with a USB RGB controller chip.

If you haven't noticed, RGB controller chips are a lot more electronically simple than that. They are small-project simple I/O-programmed chips running on simple control buses clearly designed and implemented in a hardware world of Arduinos and Raspis, GPIO buses, I2C, the old stuff. If you were around in the 1990s, you would recognize this immediately for what it is: the way things were done before NT, when DOS and Windows software frequently shoved the OS aside and ran on bare metal. In the late 90s as NT crept toward the home PC world, the impending loss of direct hardware access was one of the obvious compatibility issues visible on the horizon with the NT world-to-come, and the begrudging acceptance that this stuff was all "legacy" and we'll all have to drive our hardware like grownups in the future began.

There is a USB driver designed for simple things. It's called USB HID. And while "Human Interface Device" is actually very misleading because it's really a way to send things via USB that can be described as a report. And USB HID is a standard interface, because well, human interface devices use it. But also things like UPSes and anything else.

You want RGB control? HID provides a stupidly simple way to do it, and you don't need anything high speed.

Alternatively there are plenty of I2C busses on a PC - the modern chipset provides easily a half dozen serial-bus controllers which can be made into UARTs, I2C, SPI or other devices. You don't have to hang it off SMBus. It's just a minor BIOS tweak to initialize the chipset serial I/O controller to the right setting. The reason manufacturers don't like this option is they have to write a driver and getting it WHQL signed and all that. But the Intel and AMD chipsets likely have reference drivers.

If you're worried about costs, I'm sure there already is a USB RGB chip from some Chinese supplier. And likely a half dozen RPi Pico versions.

And likely why bother with RGB. There are USB to WSxxx addressible LED chips out there so you can control those addressible pixel LEDs from your computer

Heck, there are ESP32 projects where you can control those LED arrays over the network so you can have your room and PC dance the same 2way.

Oh and one more thing. Walmart could pull a "Winco Foods" and not accept Credit Card payments at all. In my area there's a Winco Foods about 30 miles from where I live. They only take debit cards, not credit cards. They do beat Walmart's prices so whenever I'm nearby, I take advantage of that.

Yes they could, but they'd lose out on a lot of money. Winco foods likely saw their average receipt size and realized most of their patrons could either do debit or cash. Being a grocery store, cash is likely a viable option if you're not doing big shopping trips and spending $20-40 a trip, and more for debit.

Walmart however does sell more expensive stuff that most people will not have cash for which can be expensive. Think video game consoles, TVs, etc. Limiting yourself to debit and cast would make those transactions very annoying.

It's why Costco always retains a credit card option - because they know cash and debit only get you so far. (And because debit protections are worse, I refuse to use debit, so it'll be cash only).

Sorry, no, any touch screen interaction is requiring you to look at what you are pointing at.

Which is fine when they're things you shouldn't be touching while the vehicle is in motion.

Other controls, like HVAC, radio volume and the phone answer button have to be real buttons you can access without looking. While the car is in motion, the touchscreen should disable itself and be a view-only thing - showing things like where you are on a map or what your next turn is. Anything you might need to adjust quickly should be a knob or button.

That means changing the radio station or media source - you do it while parked (it's not an emergency if your least favorite song comes on). But if you need to turn down the volume that is something you might need. As with answering a phone call. Texts and other things, not important enough to be done while in motion and if they text you, they aren't expecting an immediate response.

Ditto HVAC controls - defogger, A/C, heating, fan speeds and temperatures are stuff that need adjustment while moving to be comfortable whilst driving.

Nothing wrong with a touch screen. Just as long as the UX designers take into account that the screen should be view only while the car is in motion, and anything you might want to adjust must be available on a button. You shouldn't need to use the touch screen while driving.

And a large screen is nice - it can display a huge map, show driving information and other things without crowding out things. But it shouldn't need to be interacted with whilst in motion. Heck, why not toss in a couple of unprogrammed buttons so you can have the option of a few more adjustments you can program on your own. Maybe you might want bass boost or something. Or a quick "everything off" button that turns off climate control, radio, etc.

USE FUCKING FIBER.

Not hard, either. Because there's a cable you can use - the poewr lines. Fiber lines can go along with power lines quite neatly, and many power companies already have fiber network deployments. In fact, there are more than a few rural utilities where your electricity provider will also do fiber. (Typically GPON).

The power cable is being run to the house anyways, and the poewr company's portion is the triplex to the weatherhead. From there it can run down the weatherhead and end at the ONT. Fiber cables cannot carry power so running them along power cables is perfectly safe.

Mostly in the inaccuracy.

You can tell at best a "blob" walked between two endpoints that are participating in the monitoring network. You can't tell what that blob is, or even how big the blob is - you can guess based on how much distortion was caused, but telling the difference between a toddler and a dog would be impossible.

It requires a transmitter and a receiver to work. If you don't want to participate, all you have to do is not use their equipment. The gateway by itself is useless - you need their mesh network nodes and other stuff as well. And it only covers the path between the devices.

If you wanted to defeat this, you can either buy your own equipment - use your own mesh network nodes for example, or position it so the detection area just isn't somewhere you're likely to be (e.g., putting the mesh node on top of the gateway on the floor above).

Sure they could use better technology, but it relies on the body interfering with the wifi signal and better technology costs more money. It also requires a receiver capable of doing the requisite measurements and the monitored area is between the two devices. So it's basically trivially easy to defeat

People keep assuming the AI funnel is taking all their ram when the sad fact is that memory manufacturers simply stopped making desktop memory to use the raw materials on purpose built chips for data centers.

Well, it's not purpose built memory - it's HBM (high bandwidth memory), which is a standard thing AMD was heavily pushing for their GPUs.

But the fabs are being redirected from producing standard DDR5 memory to HBM memory because the AI companies are literally throwing money at the RAM companies "take my money!".

You know those eBay auctions you see where someone puts something up for sale at a ridiculous price? You know, someone wanting to sell it for $100 whilst they go for $50 normally? They don't sell and just sit around. But the AI companies are going "I need it now, I'll pay $100!". And prices rise because everyone sees those sales happening at those ridiculous prices.

The whole term "base load" was invented to justify inflexible nuclear power plants. It is not special and the trifecta of wind, sun and batteries can provide it just fine.

Exactly. "Base Load" is a fallacy. Nuclear suffers the exact same problem as wind and solar. In fact, the industry has a term for it - it's called "non-dispatchable".

The term means a power source whose output cannot be synced to a load. If you turn on a light, the power plant needs to supply extra power for that light. And wind, solar and nuclear cannot do that.

Wind and solar are obvious why, but nuclear is also problematic because it takes hours to change its output.

Dispatchable sources include fossil fuel (natural gas especially), hydro, heliostat solar, as well as storage technologies (grid battery). These can vary their output in minutes which as far as the grid is concerned, sufficiently fast. Battery is especially fast and can react in under a second which can fix sudden grid lopsidedness.

Nuclear advocates want it, and it makes sense. But it's just another source of power that cannot be controlled. Grid batteries make nuclear more practical, the same way they make solar and wind more practical. Because it's actually a disaster if grid demand drops below what a nuclear plant is producing - you cannot shut down a nuclear plant and a grid with too much production will also shut down to protect itself.

Of course, the other problem with nuclear is heatwaves, as we've seen Europe struggle with production through their heatwaves when cooling water that is too warm to use, jellyfish, and other things. And remember, European nuclear power is far more modern than US nuclear so it's not old designs or anything.

So the US can have more nuclear power, that's no question as its deployment is far lower than Europe, but it's no panacea because the US can suffer with nuclear power the same way Europe does.

I find it interesting that in the end it was for-profit businesses that helped Linux hit the mainstream.

What do you mean? Because this has always been the case. Most of the kernel developers work for for-profit companies - some of the biggest kernel contributors include companies like Microsoft, Intel, AMD, Nvidia and others.

Many projects are sponsored by companies like RedHat (now IBM, but IBM did a lot before acquiring RedHat).

There are for-profit Linux distributions as well - Canonical for example has Ubuntu, a Linux distribution you might have heard of.

And Linux is mainstream in the telecom space - Android is maintained by Google and given for free.

The whole Linux ecosystem is basically being developed by people working for for-profit companies.

In fact, it is so corporate the few times you run into a library maintained by someone in their spare time it's generally highly noteworthy - usually because the person behind it wishes to retire from the project. You'd generally assume that most of the libraries were really already maintained by someone in some company somewhere.

If you have to rely on AI to do that, all hope is lost. If you cannot take the output of AI and then fix the error yourself in the original document.

Are people so damn lazy that they're helpless if AI doesn't help them go to the bathroom nowadays?

If you're so worried about being caught out, you can still use AI in meaningful ways. Sure they cost more work than being lazy, but that isn't seen as a bad thing.

You want AI to proofread your work? Go for it. But instead of asking it to output a fixed copy, ask it for criticisms and suggestions. Then fix the document yourself using those suggestions. If you are doing your job, you probably have more context and can make it work.

The people who are scared the ones doing the scamming. Like a lawyer who files a brief written by AI who doesn't take 5 minutes to verify it. Probably takes them 15 minutes, but they charged the client a full hour for it and profiting off the charged 45 minutes of work that wasn't done.

Use AI, it's a technology that's not going away. But use it smartly and watch what happens to the output. And maybe apply some human effort to it and maintain a pipeline where human content comes in, and human content comes out. Ai can enter the pipeline, but the original document is only touched by humans who operate under suggestions made by AI but otherwise have full agency.

My resume was human generated. It was run through ChatGPT for suggestions on improvement. It made some, including helping me reword a troublesome bullet point. But I took that and edited the original document myself, throwing away the rest of the content which contained so much hallucinated junk I would lose any job I applied for.

If you're so worried about being found out, you're using AI wrong.

Will anyone even be able to get games that are 10 years old or so in 10 years? That's how you save money on games.

Generally, yes. This is one area where a physical game might not be available or get extremely rare. If it was a game that did well, then there will be reprints (all those "game of the year" or "classic" editions), but if a game didn't do as well, they often get heavily discounted and then destroyed. Though they usually only get a really small print run.

The question of whether 10 year old games will be cheaper, is a completely different question, however.

Digital games cost basically nothing to keep online, so if there are no licensing issues they generally stay online for ages.

This would also mean the PS6 would be a console I wouldn't need to get first day - because any game I would want to play will be available to purchase later on. There's no need to scramble and find a physical copy that sells out in minutes as they do nowadays - if it's digital, it's forever, so the rush to buy a game on day 1 is basically gone. Might as well wait for a game to go on sale.

It's a policy I apply for PC games - I generally wait until a game gets below $10 with few exceptions. Sure it means I play the game a few years after it comes out, but it works remarkably well enough. I mean, why pay $70 when a few years later I can pay $10. It's also not like I'm hurting for content since Steam has been around for over 20 years now. It's one of the biggest advantages of digital.

I only expect to pay maybe full price for 1 game on each platform every year. Everything else I expect to buy to cost $10. Maybe $20 if I really wanted it.

Actually, the company they used was Open Source Storage to hold the backups. OSS used Iron Mountain for their storage provider (Iron Mountain provides both physical document storage, but they also have a data center in their storage cavern to serve as long term data archival in digital format.)

OSS just up and disappeared on them suddenly (they did have a 30 day agreement to cancel). PBS is suing Iron Mountain to basically get at OSS' stored data so they could recover their archives.

Basically, by filing the lawsuit, it block Iron Mountain from deleting the data while the ownership gets resolved and means as long as the data still exists, Iron Mountain is responsible for keeping it so they can't just delete it.