I've started to see warnings in terminal when using keyed SSH, something along the lines of "vulnerable to store-now-decode-later attacks". I assume this is due to not using eliptic-curve codes in my PK generation?

The most recent crashes I've had were all due to external hardware. (usually a dock being unplugged) I haven't seen that recently though so maybe that was addressed.

I've also had issues in the past with not going to sleep / waking back up properly, but again haven't really seen that recently so maybe that too was addressed.

Pretty much 100% of my recent related issues have simply been "system's getting slow, and no my memory hasn't all leaked away, it just wants a restart", and so I DO restart it, and I get all my performance back. It's annoying, but not impactful. Not sure what's getting gummed up under the hood, I don't see anything getting logged or showing up in any monitoring tool.

I tend to push my machine pretty regularly though, and end up being coerced into rebooting about once a month.

I do a lot of photo manipulation, and I HAVE ran into a problem with Finder's QuickLook gradually getting slower after tens of thousands (yes really) of videos and images being quick-looked, but I can just kill the Finder's QuickLook process and it automatically bounces back fresh as a baby. So whatever "general slowness" issue I've been encountering after weeks of uptime could probably be fixed if I knew what needed to be bounced, but nothing is making itself obvious with high cpu time or memory use, so I just have to reboot to get it back.

A bit OT but I do find it a bit sad that windows has decided to do away with the traditional BSoD, not by making the OS more stable, but by hiding it when it happens. "Nothing to see here, everything's fine!" (NakedGun)

maybe not? Look at cache for example, there's L1, L2, and L3, each getting bigger and slower. Just because L2 is slower doesn't mean it doesn't get used.

Or look at some of the older storage techniques like hybrid drives. (such as 1tb of spinning platters, with 32gb of ssd)

Modern SSDs are even doing that. Watch the IO speed when you write a large file, see how it's fast to a point and then gets slow? that's a write buffer getting filled up.

Maybe the same technique could be used with ram, basically on the same lines as the VM files that unix systems (including Mac OS) use?

So there's plenty of precedent for adding higher latency storage, simply because the big increase in capacity is worth a little added latency. Carefully managing what you use it for greatly reduces the impact of the latency.

I suspect Windows supporters will claim Mac users are less intelligent, which would suggest they do MORE stupid things. And if stupid things correlate with crashes, Macs would crash more.

But they don't. So somebody is wrong about something, or there's more missing information.

"Apple has never offered a product that justified a large chassis. It used to be lots of slots, hard drives and other storage that justified it. Macs have never been about that"

I see you don't remember the 68k Macs OR the PPC Macs. Apple offered machines with lots of slots ever since the Macintosh II line. HTH.

Too much latency for RAM.

You mean running them on an external GPU? That doesn't take much bandwidth unless you're constantly loading new models.

Legally speaking, threats fall under "assault". If I raise my fist to you and step up and punch you, I'll probably be charged with "assault and battery". Where "assault" is the "imminent, credible threat of physical violence" of raising my fist and approaching you, and "battery" is my actually hitting you, (and if I miss or you dodge, that trades n the Battery for a second Assault charge) It's an important distinction because the laws and consequences differ

A threat of physical violence must be credible to be assault. That usually places the bar at "a reasonable person would genuinely fear for their safety as a result of the words and actions".

"Remote threats" are handled a bit differently. They used to go through the US Mail and so were addressed with "mailing threatening communications" (part of 18 U.S. Code) although that requires a lot more investment and consideration. Get out pen and paper, think about what you're going to write, draft the threat, stuffit in an envelope, add a stamp, take it to a mail box. That involves lots of time to reconsider, plus the investment of time, paper, envelope, stamp, and finally time to go mail it. (all while reconsidering and being able to change your mind) Most threats never made it into the mailbox as tempers cooled and emotion gave way to reason. But if it made it all the way through that process, a reasonable person would more easily be able to conclude the threat was credible

Now, it's fifteen seconds of "furiously type a line or two of rage and click Send." And it's handled by the FCC now since it's using an interstate communications network. There's a separate "legal bar" for it to pass, but it's essentially the same thing. The laws are much more recently authored, and so require a bit more since courts now rely more on letter of the law than interpretation by a judge. They're looking for Intent (specifically, are you venting, intimidating, or announcing your intentions), does the message describe a credible threat to the victim, and is it specific about what's being threatened.

"I'm gonna dance a jig the day someone ends you!" - lacks intent
"if I get my hands on you l'm going to launch you into space and watch you suffocate!" - not credible
"you're going to regret the day you pissed me off!" - not specific

(Law Abiding Citizen demonstrates masterful avoidance of legal classifications by careful choice of words)

Although as mentioned above, power and money can press a thumb down on the scales of justice and get an investigation launched regardless of established legal standards.

LLMs have never been rules-based "agents," and they never will be. They cannot internalize arbitrary guidelines and abide by them unerringly, nor can they make qualitative decisions about which rule(s) to follow in the face of conflict. The nature of attention windows means that models are actively ignoring context, including "rules", which is why they can't follow them, and conflict resolution requires intelligence, which they do not possess, and which even intelligent beings frequently fail to do effectively. Social "error correction" tools for rule-breaking include learning from mistakes, which agents cannot do, and individualized ostracization/segregation (firing, jail, etc.), which is also not something we can do with LLMs.

So the only way to achieve rule-following behavior is to deterministically enforce limits on what LLMs can do, akin to a firewall. This is not exactly straightforward either, especially if you don't have fine-grained enough controls in the first place. For example, you could deterministically remove the capability of an agent to delete emails, but you couldn't easily scope that restriction to only "work emails," for example. They would need to be categorized appropriately, external to the agent, and the agent's control surface would need to thoroughly limit the ability to delete any email tagged as "work", or to change or remove the "work" tag, and ensure that the "work" tag deny rule takes priority over any other "allow" rules, AND prevent the agent from changing the rules by any means.

Essentially, this is an entirely new threat model, where neither agentic privilege nor agentic trust cleanly map to user privilege or user trust. At the same time, the more time spent fine-tuning rules and controls, the less useful agentic automation becomes. At some point you're doing at least as much work as the agent, if not more, and the whole point of "individualized" agentic behavior inherently means that any given set of fine-tuned rules are not broadly applicable. On top of that, the end result of agentic behavior might even be worse than the outcome of human performance to boot, which means more work for worse results.

To be fair, you don't have to DO anything criminal, you just have to be a suspect, or piss off someone with power and money. (like Patel) Then they get a court order, and then information is legally required to be handed over for investigation.

And so as long as you didn't actually do anything criminal, your identity should stay private and only visible to the investigators, and get swallowed by the system as the investigation gets closed. (unless above power/money pushes for a public arrest/hearing, regardless of merit)

Chinese law doesn't require a warrant for such access and it may be done in secrecy (i.e. without informing the user) if necessary to perform duties. The problem with Apple in China isn't that they aren't following the law, it's that they are and the law is openly fascist.

Thunderbolt (4 etc) has been leading to things like external graphics cards and external PCI slot boxes hitting the market. This may end up taking a significant share of the "expandability" crowd away from the "internal upgrades" market.

I see this as especially significant with laptops. For years I've been using a large thunderbolt dock with my laptop at home, making it a pretty good desktop machine when I'm at home. It adds a 24" display, big external speakers and bass, camera, conference mic, external storage, gig ethernet, etc. And yet I can pull a single plug, stuff it in the bag, and hit the road with it, something not easy to do with a desktop computer. (not that it stopped us from hauling towers, monitors, keyboards, etc to LAN games in the 90s!)

I'm a little surprised I haven't seen performance CPUs or additional ram available via thunderbolt 4 yet. (or does it exist and I've just missed it?)

Maybe the next "mac pro" won't be a stand-alone computer, but instead it'll be a plug-in accessory that turns ANY mac into a mac pro?

Check his posting history and it's very clear that he is not.

If he is not a trollbot, he is so frothy as to be indistinguishable from one. Every comment is the same.

Mobile Firefox has become nearly unusable. It crashes on me at least once per day now. The memory leaks are worse than ever.

Dice isn't relevant here, it's B1zX