ac29

I have a SwitchBot Meter Pro CO2 that is just temp/humidity/C02. Works well and was inexpensive ($50 or so).

I've got a NDIR CO2 sensor about 6 weeks into a supposed 1 year battery life, still reporting 100% battery. Power use is very low since it only fires up the CO2 sensor every 30 mins.

Giving users ability to use apt with root privileges is pretty much game over security wise. Full root is a malicious package away

> Anthropic just provides a subscription - which Enterprise usually doesn't want you to use because everything you're submitting through that will be trained on / becomes part of their model.

My Pro account very clearly has a toggle for "Help improve our AI models: Allow the use of your chats and coding sessions to train and improve Anthropic AI models."

The inference providers are running batch sizes much larger than 10

I'm certainly willing to believe the M1 was the most energy efficient, but your data center didnt have anything faster than a laptop?

I did my first ESP32 project recently and was amazed you can get a system that starts up Micropython, then a Wifi AP, DNS, and Web Server in a second or two total and uses less than 512kB RAM. And thats with a high level programming language.

> IBM no longer owns any fabs

Per IBM: "IBM Research at Albany [...] includes more than 100,000 square feet of semiconductor fabrication space"

I guess that is technically a R&D fab not a production one, but they definitely have in house fabrication capability

Most real financial transactions need to start and end in fiat though, and that's where the costs are

A change in regulations is not a meritless argument, it's useful information

You could subscribe to Anthropic/OpenAI for the rest of your life for the cost it would take to host GLM5.2 locally - you need 1.5TB of VRAM just for the weights

> I'd rather use Sonnet than Qwen

I get this, though the pace of Chinese releases is relentless. Qwen3.7 Plus/Max (closed variants) feel notably better than Qwen3.6, and Minimax M3 is a big jump from 2.7 in capability as well. Both of these families had their previous major release less than 90 days ago.

Anthropic must have Sonnet 5 either waiting or cooking though, they said smaller and larger models than Opus were coming and we already briefly had the larger model.

If you use uv, python apps use a shared cache which helps a lot.

All current Qwen 3.7 models are closed though they have said more releases are coming

Been using zmx for a few months and love it: https://github.com/neurosnap/zmx/

Very similar, based on libghostty

> I'm curious what the downside for this speed is here

"DiffusionGemma's speedup is designed for local and low-concurrency inference. In high-QPS cloud serving, autoregressive models can be deployed to saturate compute efficiently, so DiffusionGemma's parallel decoding offers diminishing returns and can result in higher serving costs"

Most of that cost was presumably the 100,000 GPUs it contains, they didn't even build the building

According the the X post, each 150kW satellite is a bit over 2 tons, so a full year of launches would get you 75MW. Thats not much when AI datacenters are GW-class.

> in the process standardized not just electric charging but the plug we're all using today

Wat? Are you talking about NACS? If so, a minority of non-Tesla EVs currently on the road use it in the US, and AFAIK zero outside of North America.

And even in the US, the vast majority of EV charging stations are AC and use J1772, a SAE standard that predates Tesla's existence.

Any sane index will float adjust their holdings

The M1 isnt particularly good at inference, so pretty much every major current competitor with a 256+ bit unified memory system is better: AMD Strix Halo, NVIDIA DGX Spark, possibly Intel Panther Lake

Not true for Vanguard's total US stock market fund (VTSAX/VTI), the largest total US stock market fund in the world. Their CRSP index only requires 20 trading days post IPO, or 5 for large caps (this has been true for many years, this is not a recent change)

None of those settings set the speculative decoder to accept 100% of drafted token. I assume you are looking at --draft-p-min 0.0, if so, you are misunderstanding what it does.

More interesting than that to me is "we’re working on developing and releasing models that provide many of the same capabilities as Opus at a lower cost"

Sonnet and Haiku look real outclassed for the price with current Chinese competition.

The spark could have caused an explosion.

> Getting Laptops, both Framework and Dell are now on board, tested on Omarchy so they work out of the box. And so many other tiny things that just make the experience better. I say better but to most consumer, those are expected in the first place. And this "expectation" people have been waiting for more than a decade.

As a fan of boring Dell laptops/desktops and owner of many, I can tell you they have been well supported in every distro I have tried (Debian, Fedora, Arch, SUSE)

Kimi 2.6 is great. Qwen3.7-max benchmarks similarly but I havent used it yet

> The economy is in the gutters

Consumer sentiment is in the gutters certainly. But objective measures of the economy like unemployment and real wages look good to excellent

https://fred.stlouisfed.org/series/UNRATE

https://fred.stlouisfed.org/series/LES1252881600Q

> Everyone's been talking about breakthroughs for batteries for years.

Lithium iron phosphate has quietly gotten price competitive with lead acid and its wildly better tech. Not particularly sexy but its having a real world impact (LFP is commonly used for solar storage among many other uses).

> they unload the shares on the public market when they reach the top of their growth curve, leaving the public holding the bag

There are definitely some dogs that IPOd and went straight down, but investing in the broad stock market has absolutely not been a bag holding experience in the past decade+