Why the hell bother using an AI when you could just type what you would prompt into the comment directly?

To the point you put into the LLM, it's perfectly plausible that someone wiring up an LLM to their email and their bank account suddenly are at huge risk of a scam working. Whatever problems exist in the technical implementations the fact remains that folks greasing the wheels for their money to go out autonomously incur the risk of..... their money going out autonomously...

Your best engineers will be AuDHD, and AuDHD folk don't do well in these sorts of synthetic tests.

I can code in 20+ languages for over 12 different OS', write code that's good enough for CERN to run for particle accelerator data collection for a decade, have accelerated systems by an order of magnitude by understanding the poem space better than their previous engineers, but I am abysmal at exam style settings.

The software industry is riddled with SEs who write abysmal code but can pass a coding test easily. They're the folk that you have always needed to reject but didn't because they ticked a box.

Well the answer is of course they are out of touch and I don't actually have a 'right question', I just wanted to fit in.

No, it's just the sort of thing that the UEFI design kind of does.

You load a bunch of stuff that all initializes, just in case it might be needed. One of the key optimizations in some UEFI implementations is trimming down the things that get loaded based on an early assessment of whether this or that would be relevant to the expected boot. For example, hit key to enter setup and suddenly it has to load all the things to make sure all the various menus work.

Kind of like setting up the NIC in the OS before anything asks for network access, so that every network access isn't burdened with having to set up the network access. If the driver is already loaded then a variety of potential UEFI network code could just use it (PXE, HTTP, iSCSI, various crap from vendors like pre-OS firmware download)

We will see, but *so far* Ford has not been as aggressive about that sort of thing as BMW or Chevy. The BlueCruise requires actively curated lidar maps, so at least they have a rationale for that being an ongoing subscription. The other subscriptions are similar, updating GPS maps (who needs those with phone) and data connectivity (but if you have a phone, you don't need it). So far they haven't charged for remote vehicle services which arguably could be charged for the SIM service and/or server infrastructure they have.

Note that their point is they state they had the car's onboard system also provide AC to the load side, making the cost lower to incrementally make it a backup unit.

Most solutions provide DC and a separate inverter is required to power the house.

Presumably this came from the desire to consolidate things and wanted to do the 'plug power tools into the truck' feature and the delta to get to 'car side house scale inverter' was appealing.

Which is based on Minix.

I haven't looked into this detail, but if you do a network capture, you are likely to see some ipv6 link local activity during boot even if the server is never going to pxe boot. As it boots, it loads a network stack and generally becomes reachable ahead of any actual use. It may refrain from DHCP unless you do an activity that will call for it, but likely will be reachable over ipv6 link local.

Isn't a bubble bursting almost certainly a correction in general? What would you have expected to see in a bubble bursting?

The 401k will just become 401.

While this is true, being the subject of an experiment to see what happens when you do that is not something most people would choose for themselves. It's not *irrational* to be concerned. Absence of evidence on a largely unexplored question like microplastic infiltration in human bodies is not the moral equivalent of absence of evidence in an extensively studied question like a vaccine-autism link.

In another company, the employee could *apply* for the voluntary exit, but management reserved the right to decline to accept in 'special circumstances'.

There was a lot of pressure to accept of course, but they did cover the scenario of making sure they can keep someone if they thought they could not afford it.

The key is that either way, the severance is the same. When my company last did a voluntary buyout, it was more than the standard severance to sweeten the pot. If it is equal to severance, then people would be inclined to stick it out unless they *really* were looking to leave anyway.

If you stick it out, you get normal pay until the layoffs actually happen, and if you are so unlucky to be out of work for over 39 weeks, then you can start the UI checks to get that once the severance 'runs out'.

I'll broadly agree that self-hosting even when easy is skipped by most people, but funny to use email as the shining example of a relatively easy thing to self-host, that's one of the messiest things to try to self-host. But it is easy to point out that the vast bulk of AWS/Azure business is 'stuff people don't bother to self-host' and I know a lot of folks that would rather open a web hosted paint application than just start an already installed paint application.

I do think that the 'pure AI' companies are vulnerable to being rolled over in the long term by the likes of Microsoft, Apple, and Google. The companies that in various ways control the 'onramps' to this sort of tech (e.g. my company exclusively gives money to Microsoft for access to the models, because that's what they are used to doing for so much other spend).

So the private investors will fuel profit taking with IPO sell-off and the debt load may result in the company going bankrupt and the pieces being picked up by an older tech company.

While it's true every issue has its own dimension, even if those dimensions are linearly independent, they're not necessarily orthogonal. Independent basis vectors can still have a non-zero dot product—in other words, issues which are distinct from each other can align to some degree.