This was a real bummer for anyone interested in running local LLMs. Memory bandwidth is the limiting factor for performance in inference, and the Mac unified memory architecture is one of the relatively cheaper ways to get a lot of memory rather than buying a specialist AI GPU for $5-10k. I was planning to upgrade the memory a bit further than normal on my next MBP upgrade in order to experiment with AI, but now I’m questioning whether the pro chip will be fast enough to be useful.
The Mitsubishi Hyper heat can work down to -13F,
The absolutely best resource I’ve found for heat pump research is the NEEP database which will you give you actual BTU outputs at various ambient temperature readings:
ashp.neep.org/#!/product_list/
Also worth considering a geothermal heat pump depending on your geography, as then you have a guarantee of efficiency all year round
I am conflicted on this one. On one hand, yeah they’re just a platform, and realistically these kids would just go to another messaging service instead, but it also feels like they’re asleep at the wheel when it comes to investigating user reports of abuse.
It’s sort of an all social media thing, because I’ve reported posts selling drugs on FB marketplace too and they ignored them after review.
They quote one of the families in the article reporting a drug dealers account and Snapchat taking no action for months. I’d be willing to bet moderation is an afterthought and likely understaffed for the sheer volume of content on the app.
Ring and Blink are designed differently and run different hardware. However, I would guess that some Blink devices have the same issue. I might be wrong but I think all 2.4 GHz Wi-Fi is vulnerable to deauth

Craptacular Is More Like It
The Wall Street Journal informs us that Apple silicon was easy, but cellular modems are uniquely difficult.
Daring FireballI wonder what the takeaway was? They successfully designed processors with the m1 & bionic chips, so super curious what blocked them here. Are modems harder to design than processors? Maybe it’s the RF part?
Makes sense. I think you’d be fine with pretty much any modern(post DDR4) motherboard/CPU combo these days. I feel Linux hardware support is only really shakey if you’re using a SoC without upstream patches or if you’re using brand new hardware/laptops.
With that being said if you’re running a lot of containers on one host have you looked into docker compose or kubernetes? Maybe k8s is overkill for home use, but I think both offer support to restart containers if a health check fails. With k8s you also can spread out containers across multiple physical node, so you could just add a second RPI and “double” your resources.
Yet another company doing RTO layoffs to avoid paying severance
Yeah, makes sense, the people I saw online with “successful” results were primarily white men, which I would imagine would pop up pretty frequently in a training set. That also feeds into the echo chamber of AI hype too in my opinion