But what makes this AI model unique is that it’s lightweight enough to work efficiently on a CPU, with TechCrunch saying an Apple M2 chip can run it.

An Apple M2 can run bigger, higher-precision models than this FWIW. More important than this is perhaps whether older CPUs can run it with acceptable performance.

AI models are often criticized for taking too much energy to train and operate. But lightweight LLMs, such as BitNet b1.58 2B4T, could help us run AI models locally on less powerful hardware. This could reduce our dependence on massive data centers and even give people without access to the latest processors with built-in NPUs and the most powerful GPUs to use artificial intelligence.

This is definitely relevant to my interests especially with NPU support for such models coming. Dirt cheap ARM-based PCs based on e.g. the RK3588 are shipping with small NPUs

I have a Spitz, just don’t plan to run a VPN on the device itself as the CPU is very slow and will limit max thruput.

in the USA, LiFePO4 battery cost of drop-in 12V replacements has fallen to the point where lead-acid may no longer make sense for projects like this.

Although purported interest has never been higher, all the soft white libs I know are still deep in their full-of-shit phase and not doing a single fucking thing to prepare. It’s only the far leftists and people in the center that are on the uptake thus far from my vantage point.

Crap wallet for bewildered noobs who want to pay eye-watering fees for convenience

the problem lies within the underlying protocol.

The problem lies with Gargron doing what Gargron does, implementing whatever the f he wants for “the Mastodon network” and not giving a crap how it affects the health of the overall fediverse.

Hell, this isn’t even the first time there’s been drama over Mastodon’s advisory post scopes, not by a long shot. I kinda wish I’d saved receipts from the last couple times, some highly experienced devs have chimed in in the past.

Mastodon implemented a new feature in a way that would break (in a really jarring privacy-violating-to-some-extent way) until everyone else copied their implementation exactly.

You ever notice how Gargron refers to the fediverse as “the Mastodon network?”

He’s been doing things this way since 2017 at least. At this point, any longtime observer really has no other choice but to consider the behavior deliberate.

Then when I see people asking questions of his code or how to do things he usually jumps down their throats - or completely ignores them.

And honestly the biggest thing that irked me was that I didn’t feel he gave credit to the hundreds - thousands of other people who work to make the fediverse work.

Anyone who’s ever touched the Mastodon dev process knows that Gargron is much the same, FWIW, minus getting angry in public. These days I just have to shake my head at all the bright-eyed bushy-tailed noobs updating issues on the Mastodon repo, because those of us who’ve been around since the start know exactly how far that’s gonna go in nearly all cases - and in the cases it does go anywhere, it’ll be because Gargron implemented something similar with zero discussion and no credit where credit is due.

But yeah, follow Dansup long enough and you are guaranteed to see some regrettable behavior on main.

And the main reason Bluesky can have that is because it’s not actually decentralized.

The furry engineer who writes cryptography posts is working on this. Turns out it is not so simple in practice.

https://soatok.blog/category/technology/open-source/fediverse-e2ee-project/

If any dev should be getting roasted, it’s Gargron, for his many bad decisions over the years.

This is in part because he’s in public trainwreck mode fairly often.

Congrats, you’ve arrived at the right place!

Source: I subscribed to a ton of Lemmy communities to quit Reddit, and the selfhosting ones are so active they routinely push other communities down below the fold unless I sort by new.

btw if you haven’t got into Proxmox yet, have a look at it.

I’ve run a couple single-user Pleroma and Akkoma instances for 1 to 3 years each, one of them with a lot of follows in both directions and plenty of multimedia, and it worked fine on hardware 6 generations old on a crappy Comcast plan proxied thru a $4/mo VPS.

Other platform software (Mastodon especially) consumes vastly more compute resources than the Pleroma family. I haven’t tried self-hosting Lemmy yet. YMMV.

Mastodon is also an ultra-heavyweight in terms of compute resources it consumes per daily active user served. This is one reason (among multiple) that I would never run Mastodon as an ActivityPub microblogging instance.

Asianometry is a good Youtube channel, great way to spend 20 min at a time learning something.