This is, no doubt, going to lead to some absolutely nonsensical frothing screeds about how price signals are the most sinister threat to the free market since cocainized turbo-communism.

As best I could tell that question was about the people doing the implementation; not the system.

They professed optimism that it would be done in less than three years; but that it probably would take three years to get there with an LLM; then asked how radical people were willing to be; which sounds like a relatively polite phrasing of the idea that someone who wants to make real progress is going to need to take an approach that isn't a me-too LLM attempt.

TFA says: "though as a union representing just a tiny slice of the total Google workforce, it lacked the ability to collectively bargain."; but it also indicates that these are specifically the 'AI' division types; rather than just a small slice of the workforce generally.

Given where the hype is currently; I'd assume that the bot herders would be the segment of the workforce that would be most likely to be able to exert influence. Is this group too small even for that; or is this strictly a question of a legally defined percentage for collective bargaining to kick in, in which case being a small but influential subset of the employees would make no difference because the measure is purely size?

Aside from the question of whether it's a good idea; there are also fairly massively different incentives at work.

New companies have a habit of dying, but they also have enough room for potential growth that equity compensation has a chance of turning into dramatic amounts of money. Not necessarily oligarch island money; but significant alteration of whether work is something you need to do money.

At one of the relatively mature outfits? Aside from the company's performance being down to thousands of people who mostly aren't you; it's overwhelmingly more likely that any employer stock is going to behave like a less diversified version of whatever your 401k has in it; unlikely to get wiped out, but no particularly dramatic gains are expected.

Getting more hours out of the labor units and a better chance that they'll die or be incapacitated enough to be fired before they get old and expensive enough that we'd need to find a pretext to resource action them?

Sounds like a win-win for management.

It's also about where you want to put your complexity.

If you are more or less directly interacting with a user: you will almost certainly want to be able to handle case insensitive searches. You'll probably also want to let them change the font and the text size and color and have the timestamps displayed correctly for their region and time zone.

It's just much less clear how much of that you want to make the filesystem's problem. You could drag a full timezone system into the filesystem so that timestamps could be intrinsically timezone-aware. If you wanted to. Maybe the better part of a CSS implementation so that the color, size, and font of the file name could be as directly connected to the file as the name is. Wouldn't that be awesome?

I'd be interested to know how happy or unhappy Intel is about the situation: on the one hand anything that helps keep fab utilization as close to 100% as possible is presumably a positive; and, in the case of the Lunar Lake parts, the new-hotness may or may not actually be higher margin since there's more TSMC material in there and the co-packaged RAM made dealing with DRAM prices Intel's problem rather than the OEM's problem; but on the other it can't be entirely encouraging that the customer response skews so heavily toward "we don't want your improvements; we want cheap"; and (though Intel's data alone wouldn't be able to tell us whether this is the case or not) it would be actively bad if people do care a lot about battery life, or improved integrated GPUs, but the people looking to sell those are simply not going to Intel to provide them(obviously Apple isn't; but AMD is certainly relevant in mobile again, and not just as an ultra-budget option; while MS is at least trying to make Qualcomm happen).

Given how hard Intel goes on 'system TOPS' (counting the combined computation of CPU, GPU, and NPU if present) in their 'AI PC' marketing stuff; rather than Microsoft's NPU-or-nothing focus in their "Copilot Plus PC" marketing stuff; I would imagine that Intel doesn't much care about people not caring about their NPU all that much: they've got one they can sell you; it meets MS' requirements for their little copilot sticker, so they won't lose out in any tickbox-based procurement contest; but the meaning of OEMs preferring Intel's cheaper parts could be substantially different for Intel depending on whether it's a "buyers are price sensitive; especially with the current state of global trade" thing; or an "honestly, people who care get Strix Point; but someone needs to fill the cheap seats" thing.

The part of this that is genuinely a bit surprising isn't that nobody cares about glorious 'copilot+' NPU AI PC whatever; but that Intel is apparently having a hard time selling people on the actual improvement between raptor lake and meteor lake; which is the battery life.

Performance was basically a wash; but that's the generation where Intel significantly improved the efficiency situation.

Is he honestly so high on his own supply that he buys the "people would love ads if only they were more 'relevant'; because obviously there's not actually any tension between things you are interested in seeing and things that someone will pay to have rammed into your eye" concept; or is this just one of those situations where you look at your burn rate, remaining VC cash, and complete lack of plans for profitability and make a statement that isn't secret but is intended for the investors rather than the targets?

Someone who says "humans who can complete higher-level tasks, such as creativity and judgment, should not be stuck answering emails" is definitely someone you'd want to have to communicate with.

From the unfortunate occasions when I've been forced to try to talk to them; I'm pretty sure that Redmond has already culled anyone meeting this description from public-facing communication.

Either the young men in the US abandoning college education at record rates are mega fucked; or Bloomberg managed to pick a particularly dire case.

Not that this is anywhere close to the only job one could say this of; but I'd be a bit jumpy about that job's future as well as its present.

Vehicles aren't as far along; but the trend toward increasingly elegant and material-efficient frames that are designed for the factory's welder robots more than for service, filled out by nearly expendable plastics kits, looks an awful lot like the same bifurcation into 'uneconomic to service' and 'trivial, if you have access to vendor spares, and the vendor-authorized service center will pay accordingly' that has largely happened to things like consumer appliances and electronics.

Maybe more people than I imagine make the jump to doing comparatively skilled frame surgery and sheet metal operations than I imagine; or the people nursing used cars along because vendors want to sell 100k 'light trucks' because that is where the margin is are able to pay enough to sustain the deliberately not vendor focused section of the market? If not our eager workforce entrant is going to be looking at an increasingly bad day as time goes on.

I'm so, so, glad that we'll be able to deal with Google using control of a browser to strengthen its other areas of business by finding other people who have the exact same nefarious purposes for controlling a browser to potentially take it over instead!

What's particularly depressing, though not surprising, is that this is an astonishingly, mind-blowingly, historically ignorant assessment dressed up as futurism.

The idea of 'service accounts' under which programs not directly related to particular interactive users do things, is basically as old as multi-user operating systems(probably even older if you count some rudimentary attempts to emulate user separation without actually being able to afford filesystem permissions or memory management extensive enough that we'd actually call your system 'multi-user' in a modern sense); with concerns around the fact that 'hey, the whatever daemon will accept inputs from anyone; and if someone were able to do something as the whatever daemon we wouldn't know who it was doing that' arising naturally more or less immediately thereafter.

Someone is, straight face, basically saying 'you might need to worry about your SQL server being network accessible' or 'attribution could be tricky if "httpd" shows up as having done something; since we can't rule out it behaving unexpectedly in response to specially crafted POST requests' as though it's an exotic insight; because 'AI' and 'Agentic'.

They aren't strictly wrong; in the sense that the present state of security absolutely indicates that people are nowhere near ready to handle even the problems they've had for basically 50 years so god knows they aren't going to do better with newer and more exotic problems; but if you replace "virtual employee" with "daemon" the security problems they are talking up as new and exotic are stuff that would have been pretty banal, at least among people who could afford access to real computers, sometime in the 80s. Maybe mid-90s before having some janky NT stuff configured a little too optimistically on an ISDN line or something would have been a fairly normal business problem that everyone would have; at latest?

We can only hope that they are actually somewhat more with it about their nominal area of business; or we will, indeed, need to keep very close tabs indeed on the network access and credentials of those 'virtual employees' because they will be falling over and going off the rails at a rate that makes problems with SQL injection or unsafe deserialization look like things that people have successfully learned from and don't do anymore.

It seems like it should give some people cause for concern about the quality of their employee evaluations, candidate evaluations, or both; if they need to have a formalized "don't hire people you've recently fired" policy.

That seems like the sort of outcome that would, on the whole, emerge naturally unless your organization is blithely twitching in response to terribly poor quality data.

I think you are confusing license with scale.

If you have a really niche problem you are normally really exposed to risk from your supplier because there's not necessarily enough business in solving your problem to sustain redundancy or make the niche a compelling one to stay in. If it's an OSS tool you are at risk of being left to your own devices with a last compiled version and a pile of code you may or may not be able to work with. If it's some sort of proprietary freeware or oddball license you'll be left with a last compiled version. If it's a paid proprietary thing you'll maybe left with a last compiled version; maybe a last compiled version that will be impossible to reinstall without cracking or illegal for you to use 12 months from now. Pick your poison; but don't pretend you have an OSS problem.

If you've got a common problem it's much less likely that there will just be a somebody to walk away, get hit by a bus, get bought out, whatever; and if the vendor for a common problem is small enough for that to happen it's more likely that there will be an alternative you can talk to, or someone who is taking over the existing tool(whether as a maintainer; or because it's now part of their product line).