p_l

Will they remain bleeding edge by the time they are operational?

Kubernetes vastly reduced complexity of running things for me, in part because it sets common patterns, upon which further very usable patterns grow (like the various operators).

At this point, I have a 64GB server on Hetzner that has declaratively configured HTTPS with lets encrypt, postgresql databases, storage management, and all of that including running potentially conflicting packages.

Could I run it on plain docker? Sure. But I would have to wire it up myself, or deal with various missing pieces. Instead I have k3s where I just need to remember to update the internal certificate once a year.

It made it trivial for me to deploy reproducible blocks that are also nicely integrated and automated on certain axes (DNS, LetsEncrypt, load balancer / reverse proxy, storage management, PostgreSQL databases with backups).

All of that with considerably easier to understand and debug mechanisms in my experience than doing it with old school configuration management or, even worse, manually.

TSMC also AFAIK is forbidden from setting up bleeding-edge fabs outside of Taiwan?

The funniest thing is that DSKY was, IIRC, a "temporary solution" that became the norm, and despite all the platitudes today it was apparently hated by astronauts.

That's how Windows handles it - the OS libraries are language runtime agnostic and do not export language specific symbols nor need them, and language runtimes are distributed often separately.

So your C stdlib does not have dlopen() (in fact, it's not really part of C stdlib in POSIX anyway!) but you can link with standard OS-level ABI to OS-provided runtime services like "find a library and symbol from it" - AmigaOS did something similar.

General consensus is that neither the model nor its outputs can be protected IP

And funnily enough, most laws about compiled databases might not apply, for one.

And then there's new updates related to AI that fully take out LLMs from protection.

No, some of it is disrupting funding and organizational structure, meaning that NHS regions in England and Wales became structurally less capable due to loss of pooled resources, which also increased coordination burden.

And that's before you get into high level scam/idiocies related to Tory-beloved outsourced providers in IT et al.

Until something happens that disrupts the supply chains from abroad and suddenly there's an issue

GBP is a fiat currency issued by government through government spending and destroyed by government through taxation, it does not cost the taxpayer it's at worst a bad allocation of resources or incentivization of resource allocation.

Honestly, a lot of it is related to pretty obvious attempts at dismantling the NHS by Westminster

The basics are pretty distributed, further work involved increasing lifespan at high temperatures, but also some work towards switching at least some turbine stages from separated blades into "blisks" - entire turbine stage is one crystal, including the hub.

It was split from youtrack internals, IIRC

Honestly, some of the worst aspects of SOAP were legacies of XML-RPC. Some of the remaining messy parts were often "yes, you can not implement that, but then if you ever need functionality X, Y, Z - you will be fucked without this complexity".

Similar to how we have DER in ASN.1.

The rest in my experience was often related to crappy "generate me XML interface from my crappy Java/.NET class" because enterprise software demands speed of development over everything else, even if that speed is taken away by meetings

Yes, the released, as well as final RC for 2000, were still 32bit mode.

There's a quote, possibly from Raymond Chen, about how they originally naively thought they will be able to just change size of an int, IIRC, and that backfired seriously.

That said, 64bit release of windows was supposed to happen on Alpha had Compaq not pulled the plug

RFID keys vary from utterly dumb ID-based, to hackable challenge-response, to actual NFC smartcard (very rare).

Some of that can be trivially cloned.

Compaq dropped the bombshell about canceling Alpha just before Win2k RTM, surprising both Microsoft team involved and Compaq-side team.

Microsoft continued to use 2000 on Alpha to work out bugs in 64bit support since it was the only 64bit platform they had supported that had operational hardware (support for PPC was only for 32bit), making it important bit in support of Itanium and soon later amd64 ports.

Some of the details made for Alpha support (including extended support for software like FX!32) are now backbone of x86-on-ARM support in windows ARM builds

The reverse buyout story is popular, but totally ignores that a bunch of the decisions and policies were established under Condit who was "lifer" at Boeing, and in fact said policies were complained about by engineering from both Boeing and McDonnell-Douglas side, all before Condit was forced by procurement scandal to leave his position

And all of that was initiated by Condit, who was from Boeing side

Condit would probably still set in motion the moves that enabled the rot - moving HQ, unrealistic targets for 7X7 (now known as 787), selling off portions of Boeing to skim just the cream at the top, and so on.

Not letting Boeing grab McDonnell-Douglas military projects might have stopped some of that, but it might have also accelerated it.

It has also a presence in the financial world (under brand name "InterSystems Caché"), though mainly as no-sql database until Java and other foreign wrappers stop being enough, and suddenly you're writing stored procedures (or at least their equivalent) in MUMPS

GSM-R is mainly implemented with dedicated BTSes aligned with rail line

There is however the aspect of how much a work is derivative of another work, and that's where the ultimate mixing aspect of AI models shines - as it was documented that removing or adding a specific work in some cases impacted around ~1bit - or less than the random number generator used in some of the processes.

This is also why the output of AI model can infringe - because while having copyrighted work in training set does not result in model being considered derivative, it can produce a work that is obviously a derivative - and as such can be claimed to be a derivative that harms original creator.

Just the fact that the model in general might impact some very vague notions of commercial income is not enough, it must do so through a derivative, and thus infringing, copy

Depending how recent the card is, you might need to implement UEFI & GOP instead. Might be better approach to read any open source driver code on Linux or BSD to find out how to get basic framebuffer

Unfortunately from what I understood a production line that makes "normal" digital chips generally can't make memory cells with concrete physical differences in the process pipeline

An important point in copyright infringement is that it generally applies on distribution to other parties.

So the process of acquiring inputs may or may not be an infringement, but with at least proposed EU rules it does not matter to created model itself.

The exception being that output it produces is judged similar to infringement as human output without any "transformative work" credit to model - so similar to how a human could learn a book or painting to memory and close enough reproduction from memory would be infringement, but not generally using the ideas taken from them

EU AI Act is moving towards genAI output being non-copyrightable and that you'd need to actually prove derivative character from a specific copyrighted work(s) to claim infringement.

AFAIK american law is going towards similar setup.

You can imagine many internal APIs as taking "look up this <name> from directory <handle>". Surprisingly often the only practical limit is that the name does not contain NULL bytes.

Path separators, whether to accept directory entries with path separator in them, etc. are usually handled layer above

Filesystems usually do not see path separators at all, it's something handled at VFS level