I have a use for AI, specifically OpenAI's. I have a ton of legacy code written in (of all languages) Forth. This is for embedded devices and it's complicated by control loops that are written in machine language in order to keep the performance up on these controllers with low cost industrial grade CPU's. The code is pretty slick all things considered, but it's a really hard to maintain mess mostly because I'm not a Forth programmer, and especially with the machine language stuff in there it's complicated.

About a month ago I embarked on a journey to re-code all our software in C. ChatGPT (paid) has been incredibly useful in both helping me understand what's going on in the Forth code, and helping to translate it into C. The new code usually isn't compile-ready but it's close enough that it just takes a few changes and I can compile and burn to the controller. Yes, I actually can have it translate the code directly into a compile-ready version but learning the process and learning how the code translates has been incredibly useful in helping me with the old Forth code and has also helped me find fixes and optimizations that mean the new C code runs better on the same hardware than the Forth stuff did.

So yeah... one good use for AI. There ARE good uses for it but it's a tool like any other. It's also great for help with writing, especially formal writing.

There's a part of me to this day that's still mad that nobody on CIX (English BBS with early Internet access) told me about Yggdrasil. I downloaded Slackware as my first Linux distro and yes; went through the hell of configuring that bitch.

I mean, I learned a lot too but the pain of that installation set back my Linux usage "in anger" by months. Though once I had it up and running it was a ton better than my Windows installation.

Thanks for unlocking that core memory BTW. I was trying to quit drinking this year... guess that's going out the window now...

You realize that's almost exactly what people said in the early 1980's when the Y2K bug first started to become a concern?

This!

As I recall reading, all of the bills serial numbers were recorded before the money was given to Cooper and none of those bills have ever turned up in circulation again. Most likely they are spread all over the wilderness and have probably succumbed to the elements by now.

I develop embedded systems for a living. I'm not sure why I'd need GPIO and I2S pins on my laptop directly when I can buy a simple USB to GPIO or I2S adapter. Heck, I have both hooked to my docking station so when I'm at the office I can work on them, but not sure why I'd ever need GPIO sitting at a coffee shop or on an airplane unless I feel like getting arrested for being mistaken for a terrorist building a bomb.

Not sure I fully agree with that; there have been some incredible performance increases in laptops in the last few years, and incredible efficiency improvements as well. I went from a Dell Latitude with a 9th gen i7 to a Framework with an AMD 7840U and I've been completely blown away with the performance and efficiency compared to that ~6 year old Dell. CPU and GPU performance are incredible and yes will allow me to do stuff on them that I found taxing on the Dell. For the record I do a lot of CAD and CAM work on my laptop, as well as circuit board design for embedded systems.

I'm also delighted that GPU performance improved to the point that I can happily play a huge number of even pretty recent games. Sure, not the super demanding AAA titles on ultra quality, but even some premium titles I can play quite happily at medium settings and they look and play just fine. This in a slimline 13" laptop is really quite impressive and I've been really happy with the Framework.

J1772 is typically used for 11 or 22kw stations, otherwise known as "destination chargers". These are AC and not DC chargers and are typically only running at household voltages (110 or 220 in the US) and are the same as people use to charge their cars in their garages.

The standards everyone is talking about here are for DC fast charging which are distinct from the AC chargers. Tesla's NACS adapter can carry both AC and DC voltages depending on what it's connected to; in a home or destination charger it's usually AC while at a DC fast charging station it's at home.

Adapters exist for all of these and work well. They are also easy enough to store in your car so it's not like it's a huge issue. J1772 is more common because it was a standard long before NACS, but new installations of destination chargers at hotels and the like are typically using the NACS connector so you'll see the older standards phase out over time.

My car has CCS and recently got access to Tesla Superchargers (Polestar 2). I keep a NACS-J1772 and a NACS-CCS adapter in the trunk of my car. The former allows me to use NACS destination chargers while the latter should allow me to use NACS superchargers... though because I just received it yesterday I haven't had a chance to test it. I will probably test it this weekend as there's a Tesla Supercharger about 20 miles from my house.

A lot of my old clients (I am out of the systems engineering business but still meet up with my old clients periodically) are taking the opportunity to refactor many of their applications to cloud-native solutions, both private and public clouds. It's a pretty large lift-and-shift and usually requires some staffing changes or training for the new server stacks, but it's an easy sell to the C-suite when you're looking at such a significant price hike. This is especially true for those clients with perpetual licenses who thought they had a cost-controlled infrastructure stack and now clearly don't.

Where legacy apps are still required in virtual machines I've seen a lot of them go to Hyper-V but on a much smaller scale than their previous VMware stacks due to the above migrations. In some cases, they're also shifting these VM's up to Azure or AWS as stop-gap solutions while they identify new products or migrate... I have always said there's nothing more permanent than a temporary solution but in the cost analysis moving 100 VM's out of a 2000 VM stack to Azure is still cheaper than paying VMware for the licensing.

I've also seen a big uptick in interest in Proxmox, particularly using Ceph as the "VSAN" environment. It's a VERY solid alternative and the transition for the admins is actually very little as administration is very similar to managing a VMware cluster.

I've also seen a few who have decided that monolithic solutions are dead on arrival because they decide that the companies behind those solutions can pull a Broadcom at any time. These companies have started or have already finished rolling their own solutions using OSS. Ceph as a clustered filesystem, KVM as the virtualization layer and pick your poising for administration... there's tons of solutions out there and what you need depends greatly on the complexity of your environment, the skill set of your staff and how much you expect to be caring and feeding for the infrastructure. Set up properly these solutions should mostly get out of the way to allow your organization to do actual work and just require administration when something breaks.

I'd posit that most enterprises should actually get rid of anything that doesn't support some method of automating stuff. Ansible has been a thing in the enterprise for a LONG time, SSH even longer than that. If a device doesn't support automation of any kind it seems time to pressure the vendor for one or to switch to a better product.

Can't speak for A10, but I used to manage a fleet of F5's and they have a well documented, scripted method for updating certificates using SSH with some scripting. If you can't do SSH then they also support the same process over a serial console. At least they used to a decade ago when that was my job... seems likely they still have that functionality.

Any device on your network should have a well documented method of doing remote administration, and that includes certs. If it doesn't then it's not Enterprise ready in the first place and should be nowhere near your network. I have also worked with Kemp load balancers and the manual contains a well documented API for updating certificates.

If you can manage certs via SSH or serial console, you can update certs by a script. If your device doesn't support either of these then it probably has a certificate update API that is documented in the manual (that you probably haven't read) that can be scripted. If your device has none of these then it's a shit device that was built for hobbyists, not enterprises.

You realize that even for isolated network devices there's such a thing as a console? Serial consoles have been the way to manage secure network devices for decades as even when isolated they still need SOME way of administering them. So even if you aren't allowed to attach your machine to the secure network, you can connect to the serial console and update certs from there. If your device doesn't support that then you DO need a new device.

I don't disagree there is a level of burden here but in a secure network environment wouldn't you WANT the better security of shorter cert times?

Besides, in my experience most secure networks use self-signed certs and the client and server devices in that network have the root cert installed on them for trust.

Google's grip on search slipped as their search results became nearly fucking useless.

Literally just last night I had to Google "2010 audi a4 fuse box" because I was working on a car. I got FOUR FUCKING PAGES of YouTube links... as a test I hit IMAGES and still got four fucking pages of YouTube links. I don't want to watch a fucking 20 minute video about some douchebag's cat and some TV dinner by mail bullshit ad just to find out what fuses are where and even then have to pause the video at just the right moment to be able to see it. I want a fucking diagram of the fucking fusebox along with labels.

This is not a unique situation either.. the enshittification of Google search has been going on for a long time and now they're just deliberately surfacing all their own content first instead of actually surfacing relevant content.

Had the same problem; Pat Gelsinger is absolutely an engineer (and a good one by all accounts). You're right though in that Intel had years of being run by "career CEO's" rather than technology people.

In this case though he didn't get this long of a sentence for one charge; this was multiple crimes each with their own charge and own sentence. The 24 years is literally just the aggregate of the sentence for each charge.

Now that the judgment has been passed down the lawyers begin the negotiation. While what this guy did was heinous he won't serve these sentences consecutively but more likely some mix of serving his sentences simultaneously with some additions for the number of crimes and some "no early release" caveats that mean he will probably serve exactly the 7-10 years you're suggesting here.

This isn't TV or a movie where there's only one single "guilty" verdict and the judge hands down a sentence where the defendant is dragged off screaming about how he was framed. Sentencing is the start of the negotiation phase, not the end.