ranger_danger

Their github account lists over 5,000 commits and 1,500 pull requests... just in the last 4 days.

Was said Claude AI prompted to even attempt to look for anything "innovative" or just make more of the same? The same can be said of humans if one isn't already intentionally trying to do something different.

> But proper original thought is still firmly in the realm of humans

Copying isn't just how design works, it's how everything works. Humans are imitation machines.

We create new things by collecting, regurgitating and mutating stuff we experience, just like LLMs. In a vacuum man has no ideas outside of base impulses.

Hence why originality is a novice belief. The closer you get to any field, the more you realize the stories around who made all the breakthroughs are BS media narratives. Most if not all steps forward in any field have hundreds of people clawing at similar ideas concurrently.

Have you actually used Bonsai 2 though and not just the original Bonsai? The experience is vastly improved but still requires a custom llama.cpp fork to use as of right now.

What about Bonsai 2? You can fit Qwen3.8 27B on an 8GB GPU with it, and upstream llama.cpp support is already being worked on (they just got System1 support too).

This is really cool. Is the source available?

Yes, but that shifts the trust even moreso onto Apple, and now they know every site you visit AND can see all your unencrypted traffic from their servers, and can associate that traffic across multiple devices/accounts/IPs with it.

Collusion between the relay and gateway is exactly the situation they warn breaks the privacy features of OHTTP.

And while some might think "well Apple controls the whole OS, they can always spy on you"... yes, but I'm operating under the assumption that they're not currently doing this, and that collecting data at the relay level may be easier/more desirable/discrete for them.

My understanding is they can get a current location with just a subpoena (not a warrant) already, but looking at a history of past locations to watch their movements constitutes a Fourth Amendment search requiring a warrant and probable cause.

Originally the Third-party doctrine[1] also would allow searches that were otherwise illegal, but recently exceptions have been made for GPS/location history in the courts in recent years (Carpenter and Chatrie).

[1] https://en.wikipedia.org/wiki/Third-party_doctrine

What amazes me is that their software has never been updated to support flight numbers longer than 4 digits.

Does any current proxy/tunnel/VPN software support OHTTP as a transport method?

What about browsers making general website requests to services that support it?

"Suspicious" is not illegal, and neither is hosting.

Not when the definition of VPN is subjective. I could proxy/VPN through a friend's house and nobody would ever know it wasn't them.

Or just Welsh

Yes, because the amount of pirated content floating around was not 1/100th the amount that is out there now, and obtaining content is easier than ever. Especially when you have a sanctioned "library" that hosts it all for you for free.

It's about as idiot-proof as you can get IMO. In my decades of IT roles, I have come across a shocking amount of people that cannot operate a ZIP file to save their life.

Remember, these are people that cannot even find the files they have already downloaded, and then you check their downloads folder and there's 12 copies of it in there.

Note for browser UX people: The default flow for most browsers saving files doesn't actually tell you what is happening, it just... disappears into the ether as far as non-tech users are concerned. Can we fix this somehow?

> Do you think you're changing minds?

If you read their safespaces like Mastodon/Bluesky etc. where they can rail on people with impunity, yes they do seem to actually sway some people who also don't seem to mind the worst possible bad faith interpretations of his comments and just do whatever they're told.

It is used for a lot more than just finding jobs, but I think a lot of people do in fact successfully find jobs with it too, even today.

Usually what I see happen is that people will wait until they are done checking seatbelts and then take them off.

> What parent in their right mind

As a parent, you would be extremely saddened to learn that many are not.

I would be interested to see a performance comparison with qemu and softgpu (https://github.com/JHRobotics/softgpu).

I just want a locally-run LLM that I can have an actual physical audio chat with to practice my language learning, but I'm unaware of any tools that are actually easy enough to run and use.

Anyone have suggestions?

I know it's more trendy to bash them for this... but I honestly struggle a lot to understand call center people with thick accents from other countries, and I often find myself wondering "am I racist for wanting to speak to a native right now" just because I cannot make out what they are saying.

Every version of Office is still easily found online... and even if you did want to go the truly legal route for some reason for 30+ year old software, there are still many physical copies left in the world already owned by existing users and developers, or readily available on ebay etc... so I don't think it's fair at all to say it is "unavailable" or "impossible", unless you mean purely from a "buy this new from Amazon or CDW right now" type thing.

It's being worked on for upstream: https://github.com/ggml-org/llama.cpp/issues/29058

It looks like upstream work is underway: https://github.com/ggml-org/llama.cpp/issues/29058

The sheer number of geniuses working so fast on this repo never ceases to amaze me.

Please cite a reputable source for this claim, and keep in mind we are talking about Bonsai 2 and not the original Bonsai.

Even on the original one, the authors have replied to such issues with working solutions: https://github.com/PrismML-Eng/Bonsai-demo/issues/131

You might be interested in https://prismml.com/news/bonsai-2-27b

There are a few experimental solutions currently available:

https://github.com/virtio-win/kvm-guest-drivers-windows/pull...

https://github.com/qemus/qemu-vmvga

https://docs.mesa3d.org/drivers/venus.html

https://blog.getutm.app/2026/introducing-triton-directx-11-d...

If you don't need Mac or Windows guests then the existing virtio-gpu works in my experience.

You don't even need wireguard really, you could just use ssh -D by itself, then either your browser (or foxyproxy) or individual app can point to the local socks proxy, or you can use tun2socks to make a tun(4) network interface out of it. Optionally you can use a network namespace to only route certain apps through it.

Or ssh -w can do the tun(4) interface for you, and supports both layer 2 and layer 3.

True that it won't be using UDP (or sshuttle can fix TCP-in-TCP issues for you), but for my usecase it works out better anyway, as many places I frequent restrict UDP usage. Even China doesn't block SSH.

I tried Qwen3.8-27B-heretic-ara and was pleasantly surprised at how good it was. Much better than previous "abliterated" models I tried.