Overall, I wouldn't mind seeing this technology in the US, because it would cut down the lines at the relatively few chargers there are, and when RV-ing, I could "top off" the 144kWh battery which is used to run the air conditioner when camping.

I worry about stuff like that getting vandalized. Someone with a knife will just slash the coolant pipes, just because. In fact, it is amazing that vandalism of EV chargers is as low as it is, because the copper in the wires is worth something.

I'm also surprised the EU has the power grid to support these everywhere. Maybe because they are not having all their power allocated to matrix multiplication + carry (AI) or SHA256 hashes (Bitcoin mining).

I think the US car industry will likely just hike prices and add monthly subscriptions, until they are pretty much bankrupt, and then get a bailout from the government. This has worked in the past, and the execs get fat bonuses.

IMHO one of the core things the US industry seems to care about is keeping Chinese cars from getting in the market. The BYD Shark would put other full size pickups to shame for a fraction of the price. US carmakers seem to be be not interested in the bottom part of the "K" shaped economy, which creates a market vacuum, but regulation keeping players out ensures that vacuum wouldn't be filled.

A microcosm of this played out in the "B" market when people said there are too many Cadillacs and too few Chevies for RV brands, where to even buy a base model, you are looking at $150,000, and for a decent one, that is a quarter mill. The RV makers know there is a market, and other regions of the world like Australia and Europe have companies like Hymer which service the "Chevy" market. But not in the US.

I've been debating this. The absolute biggest thing in my favor for an EV is that the previous owner of my house had 240 volts, 50 amps wired up to an outlet by the driveway. This makes it very easy to just plug in a level 2 charger. Charging from home is what (IMHO) makes or breaks EVs. Mainly because in-town charging where I live sucks for the most part.

I mention this probably way too much, but I'd like to see more EREVs sold in the US. A friend of mine had an i3 with a range extender, and it was pretty much the best of both worlds. No worries about charging, just stop and get gas. If at home, or find a charger, use that for more range. There is so much improvements that can be done to make the range extender part lightweight and give better fuel economy, just because it only needs one RPM range to work effectively, rather than a large power band. To boot, the drivetrain is just an EV drivetrain which is a lot simpler.

I'm surprised EREVs are not used in heavy duty pickup trucks, where you need the insane amount of torque at low speed. This is how trains work, and even if the battery has a limited range, the simplicity of the EV drivetrain will make up for that, for ease of repairability.

This is a good point. Europe wants separation from the US. Why not fund open source projects? Everyone wins when this is done.

Unless Microsoft is moving to a new CPU architecture (Harvard architecture would be nice, but the NX bit does almost a good enough job with machines,) we are going still have hardware, the ISA, the hypervisor, the mysterious stuff that runs in ring -1 and -2 placed there by the local governments, maybe a hypervisor, the OS, then apps. Yes, we can merge an app with the OS, but even in the Apple 2 days, that was a lot of work, especially with dealing with low level I/O.

So, this means we have an OS created that has a fast path to the matrix multiplication (with carry) on the cores, with the OS as small as possible. Assuming that they will turn their noses up at BSD and Linux kernels, there is always QNX.

At the filesystem level, TernFS is what some banking industries are using at the exascale. It doesn't have permissions and such as a normal filesystem, but designed to handle data on a large scale. Might as well go with this.

For RAM, I've seen some devices that actually use the GPU's VRAM as a swap device and balloon into that.

Overall, the "AI OS" may not be true realtime, but it can help, but it needs to be able to reassign resources as need be, be it using Optane-tier storage (if it exists at all), and the OS is focused on quickly getting software requests to the cores that handle the matrix and tensor manipulation.

If it were up to me, I'd not bother writing another OS. Just using Linux and contributing mainline kernel patches would more than pay for itself, especially when the mainline patches become part of LTS distributions. If designing a hardware architecture just for AI was fundamentally so different that conventional OS kernels couldn't be used, then see about a Linux emulator so people could port tools to the OS and cross-compile. An OS needs to be able to run gcc natively, or it is not going to last long.

I'd daresay even Pages is a good alternative in some cases, and it is a free download. I prefer LibreOffice for reliability, because it works everywhere regardless of platform.

The H-1B project needs killed. Now. If someone is that important that they muscle out an American worker, they need to be given a permanent resident card with a short path to citizenship.

Surprisingly many. IUST (Iranian University of Science & Technology) wrote the SMP code for Linux in the early 1990s, IIRC. They are truly world class in their expertise.

Note -- Iranian people != IRGC.

Very true. I still have one, and still using it for a WAP which works perfectly for older devices that don't like newer APs.

I just wish Apple made them again because they do what they do extremely well.

The Neo, inexpensive as it is, still feels professional. You can tell it is a budget model when comparing it with a MacBook Air or MacBook Pro, but it definitely has a solid fit and finish, arguably on par with most PC laptops.

What I'd consider doing is maybe looking at budget models as loss leaders, and getting some upsell models. For example, I'd say a next step up would be an i5, 16 gigs of RAM, a TB SSD, with a fingerprint scanner. This way, as mentioned by another, there is some profit to be made from one group of the "K" shaped economy, while the other group, one can ship models with 64 GB of RAM, eGPU ports, etc.

There is another thing which is a money maker -- accessories, like a good docking station.

What might sell, may be a NAS. Mainly because people need backups, don't trust the cloud, so having a base station like a Time Capsule that does the job of a router, wireless AP, NAS, backup destination, and so on, would be important. People are waking up to the fact that a local server is a nice thing to have, combined with something like TailScale which means the NAS can be accessed anywhere securely, no need to open a firewall. Plus, one can sell a NAS backup/sync service so 3-2-1 protection is as simple as tossing the appliance on the network, changing the password, logging in with one's user, and then accessing shares.

This is where I find things ironic. The true activist apps were not branded as such. They were made to allow people privacy. Look how PGP was at first released. It wasn't for "activism", but keeping one's stuff private and not easily scrutinized. Back then, people were worried about governments, but the threat landscape has changed... governments don't really care, but private companies do, because they can find someone to sell that info to, or toss it in some vat to correlate with something and sell that to someone, especially in the days of being able to gamble when someone disappears or not.

What we need are apps not designed for "activism". That makes me suspicious. Instead, we need apps designed for security and personal privacy. Open source, all source code auditable, if not audited, a blessing of someone who knows what they are doing, maybe even direct government certification.

Designing for "activism" rolls the bowling ball into the gutter from the get-go. It needs to be designed to protect company data and business items just as well as the plans for a march, or time for a meet-up. It needs to be designed to have legitimate uses, otherwise, it just becomes a point where just the app's presence can be used in court to convict someone of wrongdoing.

Ultimately, with enshittification, going back to private, offline enclaves is something we may be forced to do. For example, having a PC which is offline, but is where one reads and writes messages on, and then encrypts and signs replies with GnuPG, some OpenPGP product or a simpler tool like `age` which just does public key work without the PKI/web of trust.

This will require OS vendors to be okay with offline updates... but there are items like `apt-offline` and offline ways to move patches, this may not be as bad is it could be. For example, have a patch/OS repo that is on a USB SSD, and every so often, unmount it, plug into the offline machine, have it fetch updates and reboot, then plug it back into the online PC.

I really hope GnuPG stays with us. It is one of the few encryption programs that separates the encryption of content from the transport layer. This way, the GPG message can be sent via email, a messaging app, stuffed in a file share, or similar. Its security isn't relying on TLS or anything else. Done right, it provides true E2EE, but it can take more steps to explicitly decrypt it.

I think we will see worse. There was a video of North Korea's smartphones autocorrecting any slurs on the Supreme Leader and uploading the incident. I wouldn't be surprised to see that on devices, as well as a "virus scanner" which tries to detect IP violations. Having something offline where one can decrypt messages may become a standard process for people again, if this continues.

Having been around a while, the dream of having a computer do everyone's job has been around since the 60s and 70s. Even MAD Magazine had articles and back fold in covers on this.

Companies have been wanting to get rid of workers for a while. Car makers have factories which are pretty much automated with a few guys sitting, looking down to hit the red button in case one robot gets out of sync or whatnot. In fact, a lot of production in the US (Yes, the US does make stuff) is done by robotics.

The development of LLMs just brought all that ugliness to the surface. Almost every business out there dreams of being able to shed all workers and just keep the C-levels, where if it isn't machines, it is some contractors. Even businesses that have good management are going to invest in AI and automation to keep up with the competition. Even though it may not reach promises, LLMs have been a help, provided you know how to use them, and are able to deal with hallucinations.

There are still ways to keep Moore's Law going. SMIC is getting near 2nm process nodes without EUV, for example. We also may find another process, or other materials better than doped silicon for getting near the ideal of one transistor being one atom.

We still have not expanded in the third dimension much. If we can figure out how to cool silicon and stack floorplans, this would give us a revolution. There are always some ways to do this, perhaps ionic air transfer (even though it does accumulate dust), or vapor channels.

Then, there are CPU designs and ISAs. Had Intel maintained their X86S program, this would have shaved off a significant amount of transistors needed to do everyday tasks. The age old thing of RISC over CISC may not matter now, but eventually, it may be what determines performance in the future when we can't go any smaller with nanometers, floorplans are as optimized as one can get and so on.

After that, not just CPU, but computer designs. If we went to passive backplane systems where some apps literally run on one computer and just display, while others run on another, that may be a way to do things. Pretty much a smaller version of a rack and blade system, with video wizardry to allow each blade to display apps on a screen. Add some interconnects so each backplane blade can access each other's drives via DMA and not the CPU, and this can add a few years of scaling.

I'd say, if the economy really tanks, the market will be cheap PCs, like the mini-PCs you find on Amazon that are pretty much a laptop motherboard in a small case. Those are surprisingly decent, and some can handle an eGPU.

If I were betting on a recession, I'd develop for low tier Windows and built in GPUs, and distribute the game via Steam, GOG, and the Windows Store. This way, it is accessible to virtually everyone, GOG's DRM is good enough (i.e. none), and as a game dev, I can focus on relevant content. AFAIK companies like Mob Entertainment make money hand over fist by their game chapters coming out each year.

Ideally, I'd see if I could make the game run on macOS and Linux as well, because the actual users wouldn't be as many, but the PR would be well worth it.