This is building on top of the existing core product, so the output is directly tied to the quality of their core search results being fed into the assistants. I overall really enjoy all of their A.I products, using their prompt assistant frequently for quick research tasks.
It does miss occasionally, or I feel like "that was a waste of tokens" due to a bad response or something, but overall I like supporting Kagi's current mission in the market of AI tools.
The way I understand it is that Docker with LXC allows for compute / resource sharing, where as dedicated VMs will will require passing through the entire discrete GPU. So, the VMs require a total passthrough of those Zigbees, container wouldn't?
I'm not exactly sure how the outcome would have changed here though.
I don't think this is the best indicator of trends. Chess players who aren't winning multiple tournaments a year, or aren't training a kid to win those tournaments, probably have very few sources of their (own) income. They'll take any sponsors that aren't dirt.
Twitch also has an entire category dedicated to live coding, so again, not all that weird to see codegen ads or something.
Ah no, the comment wasn't directed against the chess streamers! It was directed against the advertisers that choose that medium.
To take an extreme example, it would be weird to advertise Rolex on Twitch next to computer games. So, if "AI" is allegedly so desirable, why do it with vibe coding apps?
Live coding is another thing I don't understand the benefit of, but selling vibe coding to those people might make sense.
Your analogy doesn't seem to apply because AI is supposed to become an every-man thing or a utility, as are computer games, but a Rolex is supposed to be exclusive. The AI companies want everyone using AI. The Rolex company wants their brand to be associated to wealth.
There's a few issues that have been brought to light in the last couple years at Hackfest and other events related to LoRaWAN / Meshtastic (and derivatives). I think most notably was the failure in entropy generated during the flashing process, detailed here - https://nvd.nist.gov/vuln/detail/CVE-2025-52464
I think we're a bit past the initial AES issues, at least the Meshtastic project promptly alerted people to their crypto issues and encouraged everyone to update firmware asap.
It's not too hard to use, as long as the hardware is flashed and ready. For the end user, it's an app that connects to a bluetooth connection. I think it would very trivial to have a few good LoRaWAN ops in the community, flashing nodes en masse and handing them out to peers.
I believe it saves time, the documentation is available and easily digestible, and there's thousands of existing dotfiles to take from. I'd be disappointed to hear Claude couldn't do this.
I'm mostly pointing out that the documentation is very easy to read and implement for most tiling WMs, without the need for a coding agent.
But I'm primarily talking about the missing pieces that most tiling window managers have that you need to implement yourself, or the annoying bugs that are buried in github.
I need a lock screen; fine, hyperidle. How is it configured? Once it works it works.15 seconds with Claude or 2-3 minutes googling and implementing. Why the hell would you not use it?
QT apps have fuzzy fonts in Hyprland. Turns out that's because I was using 1.5 fractional scaling on my 4k monitor, which was information buried in some github that has barely any traffic, which Claude found while I was doing actual work.
The google meet PIP window strobes because who the hell knows why, but that too was solved by Claude finding the right github ticket and applying opacity 0.999 instead of 1.0 for that window specifically. Where is that documented in the hyprland manual?
The point is that tiling window managers _in my experience_ always have rough edges, and I've been dipping in and out of them for 20 years. Now that many people (I guess not including your good self) are using LLMs all day every day to move faster in producing code, you can apply the same tooling to bring the tiling environment up to the same level of quality that we're used to with the bigger DMs that have a lot more resources and eyes on them.
How does one feasibly purchase IP blocks these days? Most blocks that are available are starting in the six figure range, generally higher from what I see.
Six figures should get you one hell of a block, it's more like 6k for a /24 IPv4 block sale.
An IPv6 block, as used here, should be "free" (as in registration fees alone).
If you register a new company with only IPv6 there are ways to get one or two "free" (as in registration fees alone) /24 IPv4 blocks to aid in NAT64 and DNS hosting for your first ASN. All in all it was something a little over 1k for me to get an Org ID + ASN + /40 IPv6 + a /23 IPv4.
I fall into the new school gen z category, and I think you're right. We don't care. We don't care about the problems started before us, and we owe nothing to no one (but our employers, must increase value for shareholders of course).
I simply want to survive. I'll kiss ass where I have to, but not to people I don't work on behalf of.
Can't say that's entirely true for me ('02). If my [ employer, supervisor, ... ] provides me with logical, traceable tasks with their context properly laid out, I can totally put a ton of effort into providing meticulous, well thought out solutions, that are as good as it gets under the provided constraints. It's the non-sensical (be it actually non-sensical or just not understood enough because of unprovided context) tasks that make me not care.
I'll throw in my $0.02, as a fellow zoomer. I care about the things that are mine (as in, my code, my decisions, etc. etc.). But if management fucks up and tells me to fix it, there is no amount of money that will make me care. Especially if I advised management _not_ to do that in the first place.
A lot of your points are valid, however... some are taking some liberties. Are you actually running into usecases where FreeBSD is easier or faster to install than current release Fedora Server?
> The wifi thing is no problem with...
You're seriously proposing end users run Linux VMs with PCIe Passthrough to get modern networking cards to work?
A lot of wishful thinking in this thread about FreeBSD on workstations.
> A lot of your points are valid, however... some are taking some liberties. Are you actually running into usecases where FreeBSD is easier or faster to install than current release Fedora Server?
It is just that the Fedora installer is more complex... and also will fail often at partitioning or during install. I've done it hundreds of time and it failed dozens on time.
I would still recommend Fedora to Linux users but the FreeBSD installer much more simple and straightforward.
> You're seriously proposing end users run Linux VMs with PCIe Passthrough to get modern networking cards to work?
It is an Alpine running on the hypervisor you won't even notice it. It consumes less than web browser tab...
Plus it has benefits from a security point of view.
I would rather FreeBSD devs focus on other things than porting all wifi drivers.
I think the most complex part of a desktop linux install is the partitioning, but sane defaults are handed to you on most installers, so I'm curious about the failures and what induced them.
As for the whole wifi thing... Yeah man, FreeBSD isn't ready for vast majority of people, even linux veterans. I know getting the manpower to write those drivers isn't always possible, but we're talking years of this being ignored. Which has led to solutions like yours.
Something trivial to us, is not for others. It's pretty insane to even think that is a supported solution to that problem.
Personally even as a FreeBSD fanperson I wouldn't want to rely on wifibox no matter how elegant it is to use. It would forever irritate the “omg ugly hack” part of my brain lol
I installed FreeBSD 14.3 on my Framework Laptop 12 and the stock Intel AX211 Wi-Fi card Just Worked™ out of the box in FreeBSD 14.3 after a `fwget` to download the proprietary firmware blobs (removed from base between 14.2 and 14.3, FYI) while USB-tethered to my Android with a simple `dhclient ue0`:
reply