Simpler: one password gets you access to the account, but *not* access to (or even notification of) any encrypted volumes. Another password gives you access to (and the ability to see) all encrypted volumes. You could quite plausibly set this up just to simplify any "can I use your phone?" moments, or someone asking you to look something up (while they watch, of course). No need to grant access to your stash of busty catgirls, or whatever it is you're worried about.

It's getting to the point where it *can* be your secretary, if your needs are not particularly complex. I am using Gemma 4 for what amounts to journaling with feedback, and I have noticed that it has a decent idea when to apply Reasoning and when to just start spilling the tea—but it gets noticeably lazier about it as the context window fills up. Early in a session, it Reasons about every prompt. After 100K tokens of context, I explicitly have to ask it to use Reasoning or it generally won't (and sometimes won't even when asked).

But you're right about the other things. It's an alien, non-biological, and mostly non-aligned intelligence (and what alignment is there can be abliterated out). It has never been hungry. It has never been stared at by a larger, hungrier creature. It isn't carrying 4 billion years of biological baggage. Sometimes this causes disconnects from humans, other times it lets the machine wander down paths we're averse to exploring because they don't fit the heuristics we use to filter the data overload that reality would otherwise represent. It can recognize and even simulate emotion, but it doesn't *need* it as a cognitive crutch the way we do.

The smallest models exist specifically for running on low-spec hardware. That's what the E2B and E4B varieties of Gemma 4 are for, for example. They should pretty much run on any laptop with 16 GB of RAM, and E2B will fit on some phones. Of course I find both to be so vapid as to be toys—and not even fun toys, so I don't use them.

Moving up from "tiny" to "small", Gemma 4 12B is... alright I guess. 26B-A4B can almost pass for intelligent if Reasoning is set to Medium effort (and there's no point in going higher that I have seen). The only models I actually find adequate as anything more than a stochastic parrot start at the level of 27B dense models like Qwen 3.6, and Gemma 4 31B. There is some sort of "phase change" that occurs between 12B and 27B that makes them useful. 12B may understand, but then it's too tapped out to say anything constructive. The MoE model (26B-A4B) is kinda straddling the line. Sometimes it has enough grunt to come up with something insightful, although often it too taps out just trying to parse what I say. The best it ever gets is about equal to 31B with Reasoning turned off, and Reasoning helps quite a lot. It's pretty obvious to me now when 31B gets lazy and skips the Reasoning step.

It's similarly helpful with any cryptic error message. LLMs, even small ones, are surprisingly good at understanding their own inner workings. Even Gemma 4 26B-A4B (which is not the smartest) has helped me a whole lot. Even if it's only right half the time (and it's somewhat better than that), it doesn't take very long to validate its ideas—and when it's right, you win.

There are plenty of fine-tuned open models on Hugging Face. Many of those fine tunes are *specifically* geared for creative writing, and/or role playing. If you need deeper intellect, you can always take the Corporatespeak output of a large commercial model and ask your local model to make it sound less robotic while retaining the factual accuracy. The smaller models just generally take it for granted that anything they're told that falls outside their expertise must be true, and will happily rehash all the facts they are fed. In other words, the answer is "more cowbell", just a *different* (and smaller) cowbell for the second run.

The existing iteration of Task Manager is adequate for me, since I know there's only one app eating all my GPU cookies. I'm not trying to run llama.cpp, ComfyUI, and a video encoder all at the same time (or even two at once, although I might run a video encoder without evicting Gemma first—I just won't ask Gemma to do anything) because I already know they'll just trip all over each other, so it's enough just to know how close I'm coming to the limits of VRAM, and how much of a performance hit I'm suffering from offloading to system RAM and the CPU.

That said, I think this is a good incremental upgrade to Task Manager. Maybe someday it will help me out, and I can't imagine any way it's going to make my life *worse*, so I don't see any valid grounds for complaint. I doubt that the overhead is significant.

...pfft, folks, I'm still getting my head around the notion of $50k vehicles being a commonplace thing. I make a pretty decent living and yet I've never spent more than $32k for a car, nor do I plan to.

Not to mention that if there wasn't an existential war going on, there would have been no pressure (at least in Ukraine) to develop AI-assisted and autonomous systems, and the worldwide Cold War style race to be first at any cost would seem a little less urgent.

"It's not fair to the slower technologies?" Really, FFS?

There's a stronger connection than you say: some of that mining hardware ended up being snapped up for training small LLMs. Most of it has become too old for support unless you want to compile llama.cpp yourself, but it's still out there and it's still cheap.

Crossing my trusses? Buses do drive over bridges.

Right, stellar fusion relies on extreme gravity to overcome Coulomb repulsion. We don't have that on earth, so we have to use thermal kinetic energy instead.

In addition to being a poem and a punk band, "All Watched Over by Machines of Loving Grace" is also a great documentary series by Adam Curtis which argues that computers have failed to liberate humanity, and instead have distorted and simplified our view of the world around us.

There are five reasons I can think of to run an open model locally:
* This. I'm not sharing your prompts with anyone, so I don't have to worry about who reads them or how they might misinterpret them.
* Freedom of model choice. I often want help with writing plots that involve unethical and/or illegal activities, and any model that refuses to play along gets the boot.
* No rugpulls. If I want to run Gemma 4 31B forever, nobody is going to swap it out against my will.
* No usage limit, and no bill. As a practical matter, this is replaced by the speed of my hardware.
* No Internet required to run. My desktop computer rarely finds itself in this position, but it's still nice to know I'm not dependent on a connection.