This also got me on my partner's Macbook. For the longest time I couldn't figure out why I could access my local services on (Safari? I forget which one actually worked) but not on Firefox/Chrome.
The specs are underwhelming, but I could see the value to a beginner being in the software that accompanies it tightly integrated with parts kits and instructions. I'd honestly prefer to see a logic analyzer instead of a mediocre oscilloscope; I feel like the projects that most people learning want to do these days are digital, and simple logic analyzers are more amenable to being cheap while still being useful.
I feel like building an nLab would be a far more valuable learning experience then using one.
The caveat being not just as a DIY soldering kit but as a full "course" in the design and construction of it.
Its got a power supply, an MCU, analog I/O, digital I/O. Learning the theory of how to read a 100kHz analog signal is far more valuable then a device that can read a 100 kHz signal.
Yeah, I certainly have my complaints about systemd but the parent's point is undermined by the fact that cron still works. If you prefer it, carry on I doubt seriously it's going anywhere. I still do sometimes.
Nvidia's software platform for the whole Jetson series was, at least in my experience, absolutely awful on the Jetson Nano and Orin boards I worked on. Has that improved at all? I did not appreciate that the only option they provided was a full desktop version of ancient Ubuntu... and even flashing the OS image was a bizarre process.
Edit: looks like they at least have a better headless option now.
Nowadays upstream Linux with UEFI mostly works, with their out of tree drivers. I’ve managed to make it work in NixOS with the stock kernel. Look at the open embedded L4T project, they have some recipes for building that. No need to use nvidia’s kernel anymore!
Also, supposedly on the second half of 2026 they were going to be moving even more stuff out of their Jetson-specific drivers as they already do for their slightly newer chips (so you could use the standard drivers, and standard CUDA builds). Let’s see how that turns out.
They’ve got external kernel modules that they use for some hardware that is not supported upstream, namely the GPU, but also some things like Ethernet I think. Everything else is upstream though, so the situation is similar to a regular NVIDIA GPU.
Who would’ve thought that NVIDIA started upstreaming stuff once they realized how much money Linux is making them?
Yeah, one of my bigger complaints especially on the Nano was the GPU only had really limited model support (iirc, mostly tflite but maybe I'm misremembering) and it sounds like the newer ones are more normal. That and what seems from the docs to be better headless support would be major improvements. Going further to mainline distro support would make them interesting to me again.
I was always disappointed by the Nano as it was a pretty capable device, but it seemed like not many people picked it up as a platform for cool things which I always attributed to the software.
Rocky's docs are also really nice. They aren't as thorough as RedHat's, but they're much more readable and concise, and tend to be written for a less enterprise-y audience.
Content wise the RedHat docs are great, but navigating the doc has a wired feeling that is hard to describe. Everything is black and white, the page has low information density perhaps because of the line space or paragraph space; the typesetting of command line and configure examples is not clear separated from surrounding text; mouse cannot select text of the command line examples; the page top is distracting because it keeps showing and disappearing as mouse scrolls up and down. Somehow the left navigation pane is also difficult to follow, easy to get lost when trying to find a section.
The real caveat is that 538 was a Monte Carlo model, and is only as good as its inputs. "Here's what the current spread in polling numbers is *given our model and the current polling and their reported uncertainties.*" Polling uncertainties are themselves computed under certain models, and those models are subject to errors. I don't think 538 hid this, but it's a difficult caveat for people to reason about because the sorts of modeling errors that have the most influence usually represent "unknown unknowns".
Building a model for predicting the ultimate winner of a US presidential election is particularly difficult, because you are dealing with noisy input data and nonlinear effects, i.e. just a few thousand votes in a few key states can completely flip the outcome. If you then have poorly calibrated polls with a large margin of error, there is really nothing much you can do.
On the other hand, it does raise the question how valuable the 538 models for something like this really are if the outcome is a coin flip anyway.
Exactly, and correlated errors, where a polling error in one state predicts similar errors across the board.
I disagree that it's all pointless though. Most basically it's smart for campaigns to have a good model and let that inform strategy where appropriate. Since the president is a big deal other people's decisions are also impacted, and in the long run it pays to have good predictions of those chances. Also, the outcome sometimes is fairly certain and that isn't always easy to see.
I agree, it's far from pointless. The 538 model is arguably close to the best you can do considering how difficult the task is, but it's important to understand it as purely a reflection of the polling data (and 538's reliability scores for polls), and that polling data is inherently flawed. After all, there are only 2 ways to perform a perfectly accurate poll: either know the outcome a priori, or run the election. We shouldn't be too surprised when models like 538 fail to correctly predict the outcome, because that's not what they represent. It's an analytical tool for understanding the current state of polling.
Compact HF/shortwave radios with transmit capability exist, but they're pretty expensive and are generally definitely portable but not quite handheld. The biggest user of such equipment is the military, so a lot of the tech is engineered for that with civilian/amateur use as an afterthought. ICOM, Yaesu, and Xiegu are probably the best known makers, and you're looking at ~$1000 as table stakes for a modern one, though there are some slightly cheaper options.
Handheld CB radios do exist and are cheap, but I've never really used them.
Analog handhelds are still abundant, they've gotten smaller and more efficient but older ones are still basically just as good as new. IMO digital handhelds are superior, but digital protocols are much more fragmented so analog remains king as a common denominator (practically every digital handheld can do analog, too).
You only really get attention from the FCC if you interfere with some other service, especially broadcast or emergency communications frequencies. Grumpy Hams will also hunt you down sometimes, but only if you're persistent. Otherwise yeah, the FCC can't practically enforce every rule everywhere and doesn't really try. Still, get a license and be a good citizen if you want to play with this kind of thing, it's not very difficult.
Agreed. One youtuber [1] calls them Sad Hams. That has stuck with me. Linked video is the FCC's recent enforcement actions. I personally find some ham gear to be slightly less sloppy 11 meter rigs and a number of others agree.
I always thought WASTE was so cool, but my family was pretty late get off of dialup to be able to actually use it. Sounds like it's still around in some form, do people actually use it nowadays?
reply