Anyone who can piss away a billion dollars on a B-movie trash fest like Rings of Power that was clearly written by a teenage power-fantasy fanboy isn't broke. Well, maybe wasn't broke before that.

For streaming services, not even that. It's a fool-me-once issue.

When I already pay for your service, and then you ask me to pay AGAIN for not having ads, that's a type of protection racket you are running, not a legit business.

Oh yeah, and the TV UI for all streaming services I've seen so far fucking SUCKS. In capital letters. It's aweful. We've had better UIs for 30+ years. The only reason I can imagine these get out the door is that the entire UI/UX team is permanently unavailable due to collective seppuku.

So yes, if your service is shitty and you make it worse by adding ads, and then ask for prote^H^H^extra service fees to remove the things you added just so people need to pay for removing it again - you seriously expect to be treated like an honest business partner?

Absolutely true. Reduced (or in many cases halted) investment in business growth is likely the primary driver of the poor economic numbers we're seeing. It's probably also helping to reduce inflation, though that effect seems to be offset by the early impact of the tariffs. As the tariffs bite harder I expect inflation to go up significantly even with the slowdown in business. If Trump manages to force Powell out early and replace him with an ideologue who lowers interest rates, inflation will really skyrocket.

Monetary inflation and price inflation are different things, but monetary inflation (which is a measure of monetary supply) is basically irrelevant to anyone but a macroeconomist, and both CustomBuild and I are talking about price inflation. Monetary inflation is a red herring.

Food prices are not going up significantly faster than overall inflation. That would actually be strange, since food costs are a major component of the consumer price index; for food prices to to up more than overall inflation the other stuff in the CPI basket would have to go down (in relative terms), quite a lot.

That's very regional. Rents have actually declined a bit where I live, thanks to a massive construction boom. This is an area where red states do considerably better than blue ones, thanks to lighter (though still excessive, IMO) construction regulation. Nationally, housing prices are flat (inflation-adjusted) over the last six months. Also, there's basically no way Trump could affect housing prices in such a short timeframe. The effects of presidential decisions on the economy generally take 2+ years to even be measurable. Tariffs are an exception, but even tariffs take some time to have an effect, and housing prices are a long way downstream.

AFAICT, healthcare costs aren't tracked on a monthly or quarterly basis, so we don't really know how they've changed during Trump's presidency. Eventually I expect the tariffs to increase healthcare costs relative to what they would have been without the tariffs.

CustomBuild's claims that the economy is improving during Trump's presidency are false, and even if they weren't, his argument that economic improvement would make massive corruption acceptable is both offensive and sad. In fact the signs are generally negative, but let's not veer into fantasy or anecdata in the other direction either.

Though... knowing what the objective reality is may get a lot harder if Trump's strategy of firing statisticians for bad news has the predictable effect.

Ah, so you're one of those who believes it's fine for leaders to steal, as long as they do a good job otherwise? Kudos for admitting the corruption, at least.

But the jobs and inflation numbers aren't really good. We're gaining jobs, not losing them, but we're gaining them at the slowest rate since the pandemic. Inflation was trending down, but has stabilized and then ticked up the last couple of months. Also, manufacturing is down and manufacturing investment is way down, as is capex investment across all industries, and the tariffs -- which even with the "deals" reducing them are still at the highest level since the 1930s -- haven't really started to bite yet.

Fingers crossed that job creation ticks back up, inflation turns around and starts going back down (or holds steady), and that the business environment stabilizes enough for businesses to begin trying to grow again, but I'm not holding my breath. Randomly jerking the business environment one way and then another is not how you encourage a healthy economy.

LOL

It really isn't.

Because those recovery mechanisms have created account access attack vectors.

This is true for YouTube Music and other things where there are contractual limits they have to abide by. But outside of that, there is no limit on the number of devices you can have logged into a Google account.

There are really good account security reasons for this.

No, you can also set up other factors. Configure Google Authenticator (or another TOTP app; they're all the same) or, even better, get a USB or bluetooth security key. You can also generate backup codes and store them in a safe place.

All of this comes down to the simple fact that account hijacking is a huge problem, for Google as well as for users, though mostly for users, and passwords suck.

I know it's more fun to be cynical and assume it's all just BigCorp being nefarious, but it's not true. I know people in the Google account security teams and they're pulling their hair out. What they really want to do is deprecate phone numbers, too, because they're actually not a good authentication factor. But users aren't willing to use TOTP or security keys and while passkeys are great, if you lose your device, you lost your passkey. The least common denominator authenticator that provides some measure of security is the phone number.

No one should be using RSA now, even ignoring QC. RSA is slow, unwieldy and error-prone. No one who knows what they're doing uses it except in very narrow niches where it has properties that EC doesn't. Every cryptographer and cryptographic security engineer I know (including me) treats the use of RSA in protocol designs as analogous to a "code smell", a strong one. If I see a protocol design that uses RSA, it's an immediate red flag that the designer very likely doesn't know what they're doing and has probably made a bunch of mistakes that compromise security. Unless, of course, the design explains in detail why they did the weird and risky thing. Competent people will know it's weird and risky and explain their rationale for using RSA in the first few paragraphs of the doc.

However, the EC-based things people should be using are also at risk to QCs, and everyone making hardware with a lifespan of more than a few years should be moving to PQC algorithms now. At minimum, you should make sure that your cryptography-dependent designs explicitly plan for how you will migrate to PQC (including on devices in the field, if relevant). You don't have to actually move now as long as you have a clear path for moving later. But if you're, say, shipping hardware with embedded firmware verification keys, you should probably make sure that it contains a SPHINCS+ key or something and some way to enable its use in the future, even if only to bootstrap the use of some more manageable PQC algorithm.

I see no evidence that it's true. It's almost always the case that competition pushes progress faster. The only real exceptions are when the competitors are able to keep core elements of their approach secret, which isn't the case here; both IBM and Google researchers are regularly publishing most of the details of what they're doing.

This is definitively, objectively, not true, at least with respect to code quality, performance and security. Where Linux shines is in support, both in terms of available device drivers and niche-specific features, and in terms of having the broadest base of experienced users and admins. If you need a kernel and OS that will run on nearly any platform, with nearly any devices, and for which you can easily hire people who already know it, Linux is your best option. It got that way not by being supremely excellent at anything but by being reliably good enough (barely) at almost everything.

But it's really not the best kernel. In fact, if you want to look at widely-used OS kernels, I'd say that both of the other alternatives (Darwin and ntoskrnl) are technically better in important ways, and that both are less buggy. I do security, so from my perspective that's the key measure, and both are definitely better than Linux, as are many of the *BSDs.

Note that I'm not knocking Linus. It's actually rather amazing that Linux works all the places and all the ways that it does, and it's a powerful testament to Linus' ability that he's still running his project, even now that it's critical to world technology stacks, including at the biggest tech companies. Being good enough at everything is hard, and that goal is probably fundamentally incompatible with being extremely high-quality, or maximally-performant in a particular niche, or highly secure, etc.

Why not? You're going to read and review the code just as thoroughly as if you'd written it yourself. The "while in production" phrase ominously sounds like you're taking some unusual risk, but you're not. It's no different than writing new code then taking it through the normal code review and QA processes to put it in production.

I picked apart your weak arguments. But, whatever.

Unless they're holding off because of IP concerns, that doesn't make any sense to me. If the tools work well enough to be worth using on personal projects, why not use it on paid work?

This is my point. Don't use it if it wastes time or produces bad results, use it when/where it saves time. One easy way to do this is to copy your source repository and tell the coding agent to go write test cases or implement a feature or whatever and keep working on it until the code builds and passes, and while it's working you do the work yourself. When it's done (it will almost certainly finish before you), git diff the result and decide whether to use what it did or what you've done 20% of. The time investment for this is negligible. Or, what I tend to do is to set the LLM working while I catch up on email, write design docs, attend meetings, etc.

Hype or lack thereof is irrelevant. If makes you more productive, use it. If it doesn't, figure out why not. If it's because you should be using it differently, do that. If it's because the tool just sucks try a different one, or ignore it for a few months until it gets better.

It is for me. Why wouldn't it be for you? If the LLM produces code that doesn't meet my rather picky standards, I tell the LLM to fix it, or I fix it. Either way, the code is going to be basically indistinguishable from what I'd write before I send it for review. I guess it's possible that the LLM could write better code than I would write, but it definitely can't be worse than what I would write, because I won't allow that.

I've been a professional software engineer for 35 years, and been involved in hiring for all but the first two, in several different companies from tiny startups to giant corps (IBM and now Google). Maybe computer engineering is different (I doubt it), but in all that time I've never seen or heard of a company that cared about GPA, because it's a really lousy predictor of ability. Sometimes recruiters use GPA as a screening tool when they have absolutely nothing else to go on, but that's the only use of GPA I've seen.

Companies mostly want to see if you can really do the job, and the best evidence is showing that you have done the job, i.e. you have real, professional experience as a software engineer. Failing that, they want to see evidence of you having built stuff, so a Github portfolio (or, before that, a Sourceforge portfolio). The best way to get said professional experience is to do an internship or two while you're still in school. Spend your summers working (and getting paid!), so that when you graduate you have that experience to point to.

Yes, absolutely, to the internship. Meh to the GPA.

That said, I'm not surprised it's gotten a little tougher. AI tools can't replace experienced SWEs yet, but under the supervision of an experienced SWE they definitely can do most of what interns and entry-level SWEs usually do.

I don't understand what this means. How do you "dabble" and what would it mean to "pivot"?

I use AI for stuff it does well, and don't use it for stuff it doesn't do well. Knowing which is which requires investing a little time, but not much.

Why should any of that matter? If using it makes you more productive, do it. If it doesn't, don't. This isn't like a new language where using it requires a significant commitment because you'll have to maintain it in the future. The code you write with AI should look basically the same as the code you'd write without it.

Also, it's questionable how many of the civilians were really civilians, given the Japanese "Ketsugo" plan, which organized all men 15-60 and all women 17-40 into combat units, trained millions of civilians, including children, to fight with bamboo spears, farm tools, explosives, molotov cocktails and other improvised weapons and had the slogan "100 million deaths with honor", meaning that they expected the entire population would fight to the death rather than surrender to invasion.

Would that actually have happened? No one knows. What is certain is that Allied knowledge of Operation Ketsugo caused war planners to increase their estimates of Allied casualties because they assumed they'd be fighting the entire population. Experience with Japanese soldiers who consistently fought to the death validated that the Japanese culture just might be capable of that.

That was an important part of the context in which the decision to drop the bombs was made.