ac29

8 ms·

151.
by ac29 8mo ago
You can use your Claude or Gemini plan with it too for now, though Anthropic and Google have made it clear this is a ToS violation.
152.
by ac29 8mo ago
Yeah I wrote a small landlock wrapper using go-landlock to sandbox pi that works well (not public, similar projects are landrun and nono). Note that if you sandbox to literally just the working directly, pi itself wont run since pretty much
153.
by ac29 8mo ago
No idea why 'sudo pip install --break-system-packages moonshine-voice' is the recommended way to install on raspi? The authors do acknowledge this though and give a slightly too complex way to do this with uv in an example project
154.
by ac29 8mo ago
The hemp is part of the finished product, so its probably intentional they picked something so fibrous. Per the webpage: "The mycelium binds the agricultural waste together, so it can be baked into durable protective packaging"
155.
by ac29 8mo ago
> They don't "have them", they're building them. According to wikipedia Micron Fab 6 in Virginia started production in 1997 and is still operating
156.
by ac29 8mo ago
Devstral 2 is 123B parameters. Thats less than 300B, but its still much larger than the 3-14B models GP was talking about.
157.
by ac29 8mo ago
Arcee is working on that, see a blog post about their newest in progress model here: https://www.arcee.ai/blog/trinity-large Its still not fully post trained and its a non-reasoning model, but its worth keeping an eye
158.
by ac29 8mo ago
> To run Llama 3.1 8B locally, you would need a GPU with a minimum of 16 GB of VRAM, such as an NVIDIA RTX 3090 In full precision, yes. But this talaas chip uses a heavily quantized version (the article calls it "3/6 bit quant&
159.
by ac29 8mo ago
2 months of design work is fast, but how much time does fabrication, packaging, testing add? And that just gets you chips, whatever products incorporate them also need to be built and tested.
160.
by ac29 8mo ago
Drawio drawings are just XML, its possible it can generate that directly
161.
by ac29 8mo ago
Especially how generous Anthropic has been recently with subscribers - extra usage in December, $50 credit earlier this month, $20 subscription getting access to Opus. It suggests to me Anthropic is less concerned with the financial impact
162.
by ac29 8mo ago
Being 3-6 months behind SOTA may be less of a hindrance than not being able to use the tools you want, though.
163.
by ac29 8mo ago
The good enough alternative models are here or will be soon, depending on your definition of good enough. MiniMax-M2.5 looks really competitive and its a tenth of the cost of Sonnet-4.6 (they also have subscriptions). Running locally is goi
164.
by ac29 8mo ago
"A rate that sustains their business" at the moment probably looks like API pricing or maybe even higher. That means subscriptions get significantly more expensive and/or limited, which is maybe where things are headed.
165.
by ac29 8mo ago
> It (CC) does have a /models command, you can still decide to route everything to Opus if you just want to burn tokens I guess it's not default so most wouldn't Opus is claude code's default model as of sometime rece
166.
by ac29 8mo ago
You can max it out via first party clients only. I dont like it either, but its not an unreasonable restriction.
167.
by ac29 8mo ago
The usage limit on your $20/month subscription is not $20 of API tokens (if it was, why subscribe?). Its much much higher, and you can hit the equivalent of $20 of API usage in a few days.
168.
by ac29 8mo ago
I think an X230 would be performant enough for 95% of the things I do, but a 14 year old CPU is going to have pretty terrible battery life for anything more than very light usage. And things that would be light usage on a recent PC, like wa
169.
by ac29 8mo ago
Opus is available in Pro subs as well and for the sort of things I do I rarely hit the quota.
170.
by ac29 8mo ago
Claude Code presumably, or another agent
171.
by ac29 8mo ago
> a $500 Mac Mini has memory bandwidth that you just can’t get anywhere else for the price. The cheapest new mac mini is $600 on Apple's US store. And it has a 128-bit memory interface using LPDDR5X/7500, nothing exotic. The la
172.
by ac29 9mo ago
That 6 Tbps optical link is the max per MEO satellite and they are only planning 128 of those. I imagine the end users of that are pretty much only backhaul customers, individual households/businesses would still need RF or wireline se
173.
by ac29 9mo ago
Comcast pulled this on me recently through what I can only describe as malicious bundling. Internet + shitty "security" software that only runs on their hardware + modem rental is cheaper than internet only + bring your own equipm
174.
by ac29 9mo ago
> they're trying to compare at iso-power? Yeah they are defining a "rack" as 15kW, though 3x H100 PCIe is only a bit over 1kW. So they are assuming GPUs are <10% of rack power usage which sounds suspiciously low.
175.
by ac29 9mo ago
Tesla sales peaked in 2023
176.
by ac29 9mo ago
> In an EV, that ~6 hour round-trip takes about 9 hours due to 2 hours of charging and a 60 mile detour. That's using ABRP, with an Ioniq 5 from Reston VA to Dolly Sods Wilderness and back That a fairly dead area of the country char
177.
by ac29 9mo ago
> Renault 5 EV charges with 11kW. This is the size of car I need. AC only EVs dont exist in the US market any more AFAIK. Looking at the Renault 5 models currently available in the UK market, they dont have any AC only models either (may
178.
by ac29 9mo ago
> It didn't come with a home charger at all. They're not cheap. Level 1 EVSE's are super cheap, almost all of them are under $200. They aren't fast (most are 1.44kW), but that doesnt really matter if you are parked at
179.
by ac29 9mo ago
Embeddedworks sells unlimited 750kbps service for $90/year. Its data only, no phone or SMS. When I talked to them earlier this year they said there was potential to sell other data rates though nothing was as low as $2/month.
180.
by ac29 9mo ago
Yeah I'm on Verizon (via their Visible MVNO) and its ~$23/mo for unlimited data. Zero complaints on coverage or speeds.

More ›