The Z80 doesn't tech anything, it's an architecture.

The Z80 and 6502 both teach something foundational: Fundamentally, this is all simple, a typical human being can fit the the fundamentals in their head, and all the brilliant things we have in the modern era are variations on those fundamental themes played atop each other recursively. Being able to write a simple loop that can PEEK or POKE at a memory address to make the hardware do something is a powerful feeling, and it's impossible to overstate the importance of seeing that power without needing to mentally wrestle with pointers first.

Starting with a relevant architecture like ARMv9-A tells the opposite story: this stuff is impossible to learn. Just this one chip has thousands of pages of mere reference material on how to program it, and that doesn't include the manufacturer-specific extensions, on-die peripherals, etc. Sorry, n00b, but you've come into a conversation decades late, and the best you'll be able to do is tread water.

With experience, yes, we can look at that and say, "Yeah, yeah, an ALU's an ALU, this accelerator is really just another simple computer in a distributed system, and there's nothing new under the sun; how's that coffee coming along?" But many of us got that confidence by working with complete computers where we could understand the whole thing--from how the pieces worked together to the individual registers of each piece. When the pieces get more numerous and more complex, we may not be able to see the details, but the details of the simpler system give us great intuition.

This reads a lot like:

• When big companies switch to less environmentally-friendly packing options and blame consumers for not recycling,
• When global industrial-scale polluters (or climate-politics types who fly by private jet) chide us for not saving the planet by driving less, or
• When tech companies crow about being environmentally responsible, but produce instant e-waste by making products that are difficult to service and impossible to upgrade.

I doubt I have personally generated enough digital data in my nearly half a century on this planet to accumulate the same energy footprint as a minute of ChatGPT aggregate runtime, let alone when spidering and training is amortized over the lifetime of the models.

Yes, data centers consume huge amounts of power. No, it's not your image macros to blame. They're just not profitable enough to qualify for an indulgence.

I think you mean a hacker trained his business using independently-produced assets, including those scraped from OpenAI, representative of the problem domain.

Even apart from having brands like "TikTok" and "ByteDance" as potentially part of the US Code, this is an embarrassment. It's a new Red Scare, and the sort of parenting-by-government that one party pays so much lip service against.

If you don't want "your data" (whatever that means in this context) used by ByteDance, don't use their programs!

Further, take a look at these two little excerpts from the bill:

SEC. 2. Prohibition of foreign adversary controlled applications.
(a) In general.—
(1) ...It shall be unlawful for an entity...
(A) Providing services to distribute, maintain, or update such foreign adversary controlled application (including any source code of such application)

"Source code", eh?"

(7) SOURCE CODE.—The term “source code” means the combination of text and other characters comprising the content, both viewable and nonviewable, of a software application, including any publishing language, programming language, protocol, or functional content, as well as any successor languages or protocols.

I don't know which programming languages "TikTok" is written in (presumably some subset of Java, Kotlin, Objective C, and Swift), but I doubt ByteDance invented some proprietary network protocol in place of the usual IP/TCP/HTTPS stack.

I, for one, will continue to update my compilers and network stack in defiance of Federal laws written by people who don't understand how computers work.

Microsoft won't say if those zero-days were exploited to target its products, or if the company knows either way.

So that's a yes, then--probably to both?

This won't get fixed until someone has the poorly-considered plan of spoofing their services or political ads behind an incumbent politician's office number.

So long as caller-ID can be spoofed, there's no way for the person actually receiving the calls to make a useful report on them. STIR/SHAKEN fails "open" is only useful for verifying that some intermediate provider okayed the caller-ID. A carrier willing to throw it all under the bus (what? A fly-by-night VoIP provider? Never!) can sign as many bogus calls as they like.

So long as the actual people committing the frauds (insurance scams, fake arrest warrants that can go away with gift cards, fake tech support calls, etc.) are out-of-the-country, the FCC is largely toothless. That's assuming the FCC can even chase the trail all the way to wherever the call originated and that the company who leased the number didn't just happen to have their access credentials "stolen" by some totally-unaffiliated-honest company running a boiler room. So far, the vigilantes on YouTube seem to be doing more good in shutting down these operations than the FCC has.

I originally misread the headline as:

Telly Starts Shipping Ad-Free TVs

Well, shut up and take my money! ...wait.

Telly Starts Shipping Free, Ad-Supported 4K TVs

A bit like peeling a potato and keeping only the skins, that.

What we need are browsers and services that virtually click on everything, all the time, multiple times, pretend to follow every ad, show interest in all of it.

This extensions exists, and it's called AdNauseum. I'm sure it wouldn't surprise you to learn that it was available in the Chrome "web store" until Google realized what it actually does. They then categorized it as malware and revoked the developer's signing key.

The extension does have a fairly heavy footprint in terms of CPU time and network transfer, but that's only because of how pervasive advertising dreck is.

Has Governor Parson threatened to call the FBI yet?

In seriousness, though, a private actor would face all sorts of liability from accidentally publishing that sort of PII on a public website. It'd be really nice to see a federal agency hold his state to account as severely.

You don't need a car that can handle each and every broken traffic light in rural Alabama. If you start there, it's OK if the system needs an additional road-side guidance assistance.

There are (at least) two fundamental problems with this approach:

1. 1. Practice is needed for humans to maintain their driving proficiency
2. 2. Continuity of attention is needed to correctly decide on the next course of action while driving

A system which is good enough to handle the general case of driving (clear roads, moderate traffic, working signals, and well-marked lanes) will, indeed, get most of the miles behind you. However, if that system requires driver intervention when things are sub-optimal, that means potentially surprising a driver with a challenging situation as that driver is getting progressively less proficient and possibly unaware of how the situation came to be a problem.

We know what the consumer use-case for autonomous vehicles looks like: people who want to read or snooze during a boring commute as they'd be able to on a bus. Whether the technology is sufficiently mature enough to support that use case doesn't matter; as soon as it looks like it is, people will act as though it is so.

I *HATE* JavaScript. I hate everything about it. It's the worst language in the history of programming.

What's to hate about a language where you can introduce infinity or not-a-number into calculations and not get an exception? Or a language where undefined values compare to defined values without an error? Or a language where adding two arrays neither does concatenation nor vector addition, but rather stringification and then concatenation--wrecking the two original values at the middle? That is, a language where the type-coercion rules are so brain-damaged that == was a security risk so it needed ===?

If 20 years ago somebody had told me I'd be doing this, I'd have changed careers.

Oh, but we were told this 20 years ago. Netscape Enterprise Server did server-side Javascript then. The only difference is that the industry took a long look at it and rightly laughed it off as a joke. FFS, we universally decided that J2EE was less loathsome.

Somehow, in the time that's elapsed, we became far less cynical and cautious.

Godspeed, fellow refusenik. I'll be over here in the Assembly/C/C++ briar patch. I hear it's awful. Maybe I'll get to the awful parts someday so that I can confirm that, but I haven't found them in the decades of searching.

I know I'm not the target audience for these devices because no component of them can be upgraded, but they really do look nice. The industrial design is amazing, the screen of gorgeous, the form factor is perfect, and they do promise a smooth Windows experience (whatever that fairy tale might be like)..

That's why my wife has a Surface Book 2, and we've had no end of problems with the thing.

The power supply is inadequate. It's rated below the TDP of the components in the laptop. Plugged into the wall, it will slowly discharge if you're beating the hell out of it. And, if you do that for long enough, the power supply dies. We're on #3 in about 18 months.

The "Surface port" doesn't have great physical registration between the two halves of the device. It'll eventually wiggle into a partially-seated state, with the devices in the lower half hopping on and off the system bus. The fault can be cleared if the device can get into a good state for long enough to detach and reconnect the base, but to do that, the battery charge has to be just so. Since this state sometimes makes the charge controller not want to take a charge, a reboot is often needed.

Support from Microsoft has been less than enthusiastic, and we're not even out of the warranty period!

Meanwhile, my gigantic Thinkpad cost about half as much, deigns to allow users to upgrade memory and storage, has a charger nearly stout enough to chock a wheel, and Just Keeps Going. Horses for courses, I guess.

It also requires GrubHub and the like to operate under opt-in instead of opt-out with the restaurants. They can't (legally) add a restaurant without their permission under this.

That's not in the legislation referenced by the post. Motherboard says that it is, but the legislation is linked, and it's very short.

The only opt-in referenced in the legislation is that the restaurant may ask for the customer data unless the customer has opted-into privacy protection as provided by some other statute.