No, they said Google tells the truth 90% of the time, not 10%.

How does any of what you said matter? Even if you think making $140K isn't upper middle class, the research still shows more people are going from $90k to $140k than those who are going from $90k to $40k. It still shows that more people in the middle class, however you define it, are going up in income than are going down.

These categories are just arbitrary lines used to make it easier for people to identify and understand the insights analysts are able to pull from the data. And however you slice it, the data shows a story of improvement. Not as much improvement as I'd like to say, but improvement overall. We have been seeing the same trend for 30+ years.

What this story doesn't show is the dark side of an otherwise good development. Most people don't really notice day-to-day how much better off the 1% are because there are so few of them (by definition). But when a growing number of people see 30%+ of people around them living much better lives, they begin to notice.

The smallest (and only) open weight model that gets Opus or Sonnet level coding performance is MiniMax M2.5, and you need about 512GB of VRAM for that model (with enough room for input tokens). At 128GB you are looking at Opus 4.0 / Haiku 4.5 level models like Qwen 3.5 122B-A1 at Q4 or Qwen3-Coder-Next 80B-A3B at Q8.

I think it's likely we will have small language models that specifically target coding that are at Opus 4.6 quality on 128-256GB of VRAM in the next couple years, I don't think we are there yet.

AI has been running at a big loss to get the users hooked. It was inevitable that prices would start climbing. That process is nowhere near done, running AI is expensive as hell.

Once the market starts reflecting the actual costs, you can bet the cost/benefit will not be nearly as rosy as it looks now. But some customers will already have gotten themselves between a rock and a hard place and will be sucked dry, then discarded. Those "expensive" people that are getting dumped will start looking like a bargain, but they will have already been snapped up by smarter companies by the time management that can't see past their own toes figures that out.

All of this makes me remember a short story reading assignment in the 5th grade. It was about kids growing up in a society where machines did all of the intellectual work. To them, writing was 'squiggles'. They managed to disable a filter on their "bard" (a story teller for children) and had it tell them a tale of machines ruling over Man.

Nobody expects prophesy from a 5th grade reading assignment.

The problem is that the people who could afford this and would do it are exactly the people the rest of society is better off without.

As a counterpoint, The Linux kernel and much of the userspace in various distros is done remote. It can work, even on highly collaborative projects. Like anything, some will enjoy that more than others.

Required physical equipment can be a limiting factor, of course. Though I have done firmware development from home because the dev board wasn't expensive nor is a debugger for that hardware.

AI-generated code is just this generation's version of copying code from the web or from another part of the codebase. Sometimes that person understood the code fully, and sometimes they just checked to see if the output matched what they expected.

The only uniquely dangerous thing about this recent iteration of that problem is the massive scale.

No, CEOs are trying to show their board, investors, and activist investors that they have a plan for how to take advantage of AI and can at least keep up with their competitors use of AI, if not surpass them. I work at a large enterprise (close to 50k employees) and VPs are being told that they need to find ways for AI to have an impact on their department or their leaders will find someone who can. If it isn't happening fast enough consultants are brought in to take over their department's transformative roadmap and leaders who can't keep up are relegated to being SMEs until they are eventually replaced. I'm not in the room when that message is given, but I've seen the rapid shift of VPs who were raising alarms nearly immediately turn into AI cheerleaders.

If you work for a publicly traded or VC backed company I assure you your CEO does not have a choice on whether to jump on the AI bandwagon. That's not how hype driven bubbles work.

What is sudden about this? CEOs have been doing this for over two years (at least). Duolingo and Klarna were among the first, and both of those were in Q1 2024. This is not new behavior.

There is a nicotine source that isn't at all age locked. I guess the teens will move on to these "cigarettes" you hear about from time to time.

They even come in cool Menthol.

The laws are a joke by people who apparently flunked "Hello World".

They demand a mechanism, but don't even offer guidance on what mechanism it should be. You can technically comply while having no 2 Linux installs following the same API, making it effectively useless.

A better approach would be a purely optional userspace package (perhaps call it "Californication") that returns 1 dword with the age information encoded in it. Each person installing it gets to decide what that encoding will be.

Yes, it returned 0x0BADF00D, that's the code for 18+.

Someone else might decide 0x0B00B1E5 means 18+.

Some early adopters of "Here's a complaint one, pretty please use it" included small operations like PGP. Others were small companies then, later to become large.

Not too long after, there was the whole flap around DeCSS for DVDs. The medium itself is nearly dead now, but it was individual efforts that rendered region coding largely a joke. The Chinese vendors whose DVD players didn't give a damn about region codes came second.

I can buy alcohol because I don't live in Saudi Arabia. I can have an OS that doesn't know or care how old I am because I don't live in California. That law literally doesn't apply to me. If I make a distro where I am, why should I bother with age verification at all? It's none of my business if a friend of a friend or a complete stranger decides to download it and install it on a machine in California. Not my circus, not my monkeys.

You can get fired at the drop of a hat for no reason at all. Bosses blow their top over being 1 minute late. Fill out these forms you just filled out last week and again online before you can see the doctor. IRS knows exactly what you made and what you paid in witholding, and what you owe but YOU need to compute it, better not be wrong! Don't be late! Rent and mortgage take up an ever increasing portion of your monthly income (if any). 23 calls a day, mostly scams. You have health insurance even though it was damned expensive, but somehow you still owe a heap of money you don't have after a single visit to the ER.

Meanwhile, you're getting badgered about your "credit score". If you let it get bad everything gets more expensive and it gets harder to get a job (for some reason).

Yeah, you're less likely to actually die today than years ago, but your place in life is far more precarious than it was even 20 years ago. More things demand your attention.