At least in the current climate (and quite possibly indefinitely; depending on how prudent their investments are and whether they have any uncontrollably expensive hobby) there's not really any reason for the 'AI' guys to take such a servile attitude.

If you actually need the job, or are invested in the company's mission for some reason, it's a good idea to care at least slightly about how your paycheck doesn't bounce; but that's not really the position these guys are in. Exceptionally in-demand skillset and reputation; existing net worth almost certainly already enough to at least keep them comfortable indefinitely if they feel like quitting the rat race or get fired. Why settle for doing sordid adtech if you think that, best case, your boss in the sort of dumbass who would lose billions of dollars on the idea that Second Life would totally have the GDP of western europe, because reasons, and you can keep him paying you a handsome salary and providing you with the GPU compute time and dubiously sourced datasets that you find personally interesting; and worst case, if you lose the fight, you'll just be told to go sling ads, not fired and blacklisted.

Facebook isn't running a charity; but neither are these guys. Why wouldn't they try to take what they can get? Especially when the actually-profitable business units are fat enough that there's plenty of room for boondoggles, so long as you can sell them, rather than there being fairly tight constraints on how much you can waste before the company starts bleeding out.

It would honestly be more surprising if they signed up with facebook out of a genuine willingness to do adtech swill and sordid 'engagement' hacking; rather than on the assumption that there's enough desperate dumb money sloshing around in Zuckerberg's fear of missing out on the next big thing that they can get paid to pursue their pet projects without much concern for having to deliver short term impact on the bottom line.

I assume that at least some of the tension here is that facebook hired these guys to be the hotshot golden boys of sucking less at AI; so it isn't just an it's-only-money thing. I don't know whether or not this belief is accurate; but Zuck and friends certainly hunted down and paid for the various new AI hires as though they were capable of things that in-house or more readily available alternatives are not, so the battle over where their attention will be focused is presumably being waged on the assumption that having someone else do what they aren't doing isn't really a substitute.

What I would be curious to know is why the 'build god-machine' goal isn't being treated as the obvious winner just because you can have the god machine make facebook more addictive and better at serving ads. Do they think that the AI guys are drinking the kool-aide and the only thing they'll actually be able to deliver is incremental improvements; so they want those churned directly into products? Some degree of confidence that they will eventually manage it; but fear of missing out on some sort of short term advantage means that they don't care about what is achievable in 5-10 years? Genuinely zero interest in anything except making social media more of a hellscape; so they simply don't care?

I suspect that, while it would be socially controversial, planetary colonization would be a very strong case for IVF and some population planning.

For the amount of volume/mass required to ship a single human and support them in transit and on site you could ship a lot of embryos in cryo(it's careful plumbing; but a big dewar flask kept at cryogenic temperatures is downright lightweight compared to a full life support system); and shipping embryos gives you the option of bringing massive genetic diversity, thousands to tens of thousands of genetically screened parents worth of embryos in the space a single person would require.

Unless you've got some sort of advanced growth vats you would obviously need people onsite; but instead of dealing with the probably-impossible task of keeping a tiny breeding population's gene pool in order you'd just be defrosting and gestating specimens from a much larger pool of diverse embryos as needed. Presumably you'd initially go with an all-female colony, and only start defrosting males and trying to maintain a viable natural population once you had at least high single-digit thousands to low-mid tens of thousands.

I'm sure that such an arrangement would freak some people out; and you'd probably need to do some reasonably intense social engineering to keep everyone on-mission; but in terms of efficiency of genetic diversity there's a fairly compelling case to be made.

What sort of disaster do you have in mind that would render earth less habitable than mars?

Short of unstoppable replicator nanites turning the entire crust into grey goo; or very long term issues with the sun reaching EoL that will be an issue for basically anyone in the solar system, it's honestly hard to think of ways you could break it more badly.

Plenty of possibilities that will make people deeply miserable; or cause 80+ percent of the population to die horribly; but you'll still have a planet with the right gravity, an atmosphere and magnetosphere, some sort of ecosystem(even if it's just algal scum and cockroaches); some soil that isn't riddled with perchlorates, and so on.

I was going to say something snide about how MCP is a laughably thin standard; but 'agents.md' is literally just a text file(encoding unspecified; I guess UTF8 but nobody actually says) that you put text in and hope your bot will react appropriately to. It describes the contents as 'standard markdown'; without mention of which markdown variant they mean.

Given that the whole thing is just an exercise in getting away with bots being more or less as OK-ish with poorly structured inputs as they are with anything else it's not like it would be a better 'standard' if there were a thicket of XML schema involved; but saying:

:"AGENTS.md emerged from collaborative efforts across the AI software development ecosystem, including OpenAI Codex, Amp, Jules from Google, Cursor, and Factory.

We’re committed to helping maintain and evolve this as an open format that benefits the entire developer community, regardless of which coding agent you use."

About a 'standard' which is 'put some kind of markdown, y'know, stuff you'd tell someone about your project in a text file called Agents.md' is a little grandiose.

They are no longer required to fully fund pensions:
https://en.wikipedia.org/wiki/...

So, in order to protect against possible military applications(known for their cost-sensitivity...); we are making the sale legal as long as El Presidente gets his cut? That's in character, sure; but what's the paper-thin excuse for that being a cogent policy idea?

One major difference, assuming you've got full platform support(should be the case on any server or workstation that isn't an utter joke; but can be a problem with some desktop boards that 'support' ECC in the sense that AMD didn't laser it off the way Intel does; but don't really care); is that ECC RAM can (and should) report even correctable errors; so you get considerably more warning than you do with non-ECC RAM.

If you pay no attention to error reports ECC or non-ECC are both rolling the dice; though ECC has better odds; but 'proper' ECC and Linux-EDAC support will allow you to keep an eye on worrisome events(normally with something like rasdaemon, not sure what other options and preferences there are in terms of aggregating the kernel-provided data) and, unless the RAM fails particularly dramatically and thoroughly, will give you much better odds of knowing that you have a hardware problem while that problem is still at correctable levels; so you can take appropriate action(either replacement, or on the really fancy server systems, some 'chipkill'-like arrangement where the specific piece of DRAM that is failing gets cut out of use when deeemed unreliable without having to bring the system down.

Sometimes you'd get a BSOD that was a fairly clear call to action; when the error called out something recognizable as the name of part of a driver; but that is mostly just a special case of the "did you change any hardware or update any drivers recently?" troubleshooting steps that people have been doing more or less blind since forever; admittedly slightly more helpful in cases where as far as you know the answer to those questions is 'no'; but windows update did slip you a driver update; or a change in OS behavior means that a driver that used to work is now troublesome.

Realistically, as long as the OS provides suitable support for being configured to collect actual crash dump material if you want it; it's hard to object too strongly to the idea that just rebooting fairly quickly is probably the better choice vs. trying to make the BSOD a genuinely useful debugging resource; especially given how rare it is for the person with useful debugging ability to happen to be at the console at the time of crash(rather than just an end user who is ill equipped to make sense of it; or a system that mostly does server stuff, quite likely not on actual physical hardware, where nobody has even touched the physical console in months or years; and it's more or less entirely useless to display a message there; rather than rebooting and hoping that things come up enough that management software can grab the dump files; or giving up and leaving the system in EMS so that someone can attach to that console.

What seems most depressing about this isn't the fact that the bot is stupid; but that something about 'AI' seems to have caused people who should have known better to just ignore precautions that are old, simple, and relatively obvious.

It remains unclear whether you can solve the bots being stupid problem even in principle; but it's not like computing has never dealt with actors that either need to be saved from themselves or are likely malicious before; and between running more than a few web servers, building a browser, and slapping together an OS it's not like Google doesn't have people who know that stuff on payroll who know about that sort of thing.

In this case, the bot being a moron would have been a non-issue if it had simply been confined to running shell commands inside the project directory(which is presumably under version control, so worst case you just roll back); not above it where it can hose the entire drive.

There just seems to be something cursed about 'AI' products, not sure if it's the rush to market or if mediocre people are most fascinated with the tool, that invites really sloppy, heedless, lazy, failure to care about useful, mature, relatively simple mitigations for the well known(if not particularly well understood) faults of the 'AI' behavior itself.

There's nothing that screams 'trying to con you' quite as hard as a participant in a nominal free-market scenario trying to preserve ignorance and information asymmetry.

What always puzzled me about Intel's...more peripheral...activities is that they seemed to fall into a weird, unhelpful, gap between 'doing some VC with the Xeon money; rather than just parking it in investments one notch riskier than savings accounts' and 'strategic additions to the core product'; which normally meant that the non-core stuff had limited synergies with intel systems; and had the risks associated with being a relatively minor program at a big company with a more profitable division; and thus subject to being coopted or killed at any time.

Seemed to happen both with internal projects and with acquisitions. Intel buys Altera because, um, FPGAs are cool and useful and it will 'accelerate innovation' if Intel is putting the PCIe-connected FPGA on the CPU's PCIe root complex rather than a 3rd party vendor doing it? Or something? Even at the tech demo level I'm not sure we even saw a single instance of an FPGA being put on the same package as a CPU(despite 'foveros' also being the advanced-packaging hotness that Intel assured us would make gluing IP blocks together easy and awesome). They just sort of bought them and churned them without any apparent integration. No 'FPGA with big fuck-off memory controller or PCIe root we borrowed from a xeon' type part. No 'Intel QuickAssist Technology now includes programmable FPGA blocks on select parts' CPUs or NICs. Just sort of Intel sells Altera stuff now.

On the network side, Intel just kind of did nothing with and then killed off both the internal Omni-path(good thing it didn't turn out that having an HPC focused interconnect you could run straight from your compute die would have been handy in the future...luckily NVlink never amounted to much...) and the stuff they bought from Barefoot; and at this point barely seems to ship NICs without fairly serious issues. I'm not even counting Lantiq; which they seem to have basically just spent 5 years passing on to Maxlinear with minimal effect; unless that one was somehow related to that period where they sold cable modem chipsets that really sucked. It's honestly downright weird how bad the news seems to be for anything that intel dabbles in that isn't the core business.

Not delivering on schedule is absolutely a symptom; it's just a somewhat diagnostically tricky one since the failure can come from several directions; and 'success' can be generated by gaming the system in several places, as well as by successful execution.

In the 'ideal' case things mostly happening on schedule is a good sign because it means both that the people doing the doing are productive and reliable and the people trying to plan have a decent sense(whether personally, or by knowing what they don't know and where they can get an honest assessment and doing so) of how long things are going to take; whether there's something useful that can be added or whether forcing some mythical man-month on the people already working on it would just be a burden; keeping an eye on whether there's anything in the critical path that is going to disrupt a bunch of other projects, and so on.

If you start losing your grip on the schedule, that fact alone doesn't tell you whether your execution is dysfunctional or your planners are delusional, or some combination of the two; but it's not a good sign. Unhelpfully, the relationship between how visibly the gantt charts are perturbed and how big a problem there is is non-obvious(a company whose execution is robust but whose planners live in a world of vibes-based theatre and one whose execution is dysfunctional and crumbling and whose planners are reusing estimates from the time before the rot set in might blow a roughly equal number of deadlines; despite one having mostly a fluff problem and one probably being in terminal decline); but it's never a good sign.

It seems reasonable; but also like something that should really spook the customers.

It seems to be generally accepted that junior devs start out as more of an investment than a genuine aid to productivity; so you try to pick the ones that seem sharp and with it, put some time into them, and treat them OK enough that they at least stick around long enough to become valuable and do some work for you.

If that dynamic is now being played out with someone else's bots, you are now making that investment in something that is less likely to leave, whatever as-a-service you are paying for will continue to take your money; but which is much more likely to have a price regularly and aggressively adjusted based on its perceived capabilities; and have whatever it learned from you immediately cloned out to every other customer.

Sort of a hybrid of the 'cloud' we-abstract-the-details arrangement and the 'body shop' we-provision-fungible-labor-units arrangement.

Some customers presumably won't care much; sort of the way people who use Wix because it's more professional than only having your business on Facebook don't exactly consider web design or site reliability to be relevant competencies; their choice is mostly going to be between pure off the shelf software and maybe-the-vibe-coded-stuff-is-good-enough; but if your operation depends in any way on your comparative ability to build software Amazon is basically telling you that any of the commercial offerings are actively process-mining you out of that advantage as fast as the wobbly state of the tech allows.

If the litigation is anything to go by; their "Suicide Solution" nostalgia callback colab with Sony Music Entertainment should be a hit!