zakisaad

Model weights (what is being tested here) don't inherently "access the web" when inference is running. If the model has access to a web search tool, that's a different story.

This is the reason why performant and "safe" systems languages are on the rise and being accepted into foundational areas of our operating systems (such as the kernel). When the attack defense surface is tighter (language, tooling, compiler instead of the actual code itself), the smart folks can stay at that layer, while the masses can write more code at a level of abstraction that nullifies many of these vulnerabilities by default.

Any chance you could drop the set of feeds? Would be very interested!

Turbos have entered the chat

raises hand, but point taken when painting with broad strokes

> I hate spending money on hardware so i'm living with it

This is an odd take to me, as personally I'd be more willing to spend money on well designed and made hardware, than most software.

This is not at all aligned with my experience. Do you have a source for this?

This is a wild take. All cars can perfectly drive around a track, so why would you ever want an F1 car?

This is interesting to me - why'd you go with the 5070 for your 4x build?

At first thought, they are quite skewed toward compute (vs VRAM), which is great for gamers but not so great for running LLMs.

(I run a 5070 in my desktop)

How do you define "correct" code?

I guess my intent was not to point out that it has a mechanical key, rather that the description of the key on the webpage was wrong (it is not hinged).

The mechanical key fallback pattern is standard across the industry for sure.

This was stated about the key: "Folded into the base is a mechanical backup key, a flat metal blade in a hinged housing."

I own a BYD: this is not true. The key is not hinged; rather, the entire mechanical key pulls out when a small clip is unlatched near the top of the assembly (you can see it in the CT). I assume the circular hinge-looking mechanism in the CT is just a by product of the plastic/metal weld process.

Nonetheless: very cool tech demo!

We're currently at this stage with out kids, too.

I think the staunch "no screen" mentality is a broad-stroke lever that non-technical thought leaders in the child wellness space have stuck to, and I understand where they come from.

Though, as someone who owes his livelihood to being able to tinker and experiment with technology as a child, I'm looking toward a more measured approach. I may very well set up an airgapped Linux box (Windows has come a long way since the XP days, and gone entirely the wrong way) and let my kids proverbially "have at it" - this way, they can't get stuck in big tech's psycho-loops or sucked into YouTube's colourful dopamine machine - which I think, is the entire drive behind "no screens".

I think well-measured exposure is imperative.

I've never understood how this can be limited in practice: surely as far as the carrier is concerned, all traffic from the mobile device is the same (unless there are identifiers on the traffic coming from hotspotted devices via the mobile device). Here in Australia we've never had any form of hotspot detection/segmentation - if you have a data plan, all data features work (across all carriers). I do recall lots of online chatter from the US though, especially years back when mobile data was more of a precious resource.

You could say the same about Codex (and other tooling). Opus as a model is market leading (trading blows with the greatest that OpenAI is peddling), but there will be a reckoning when open weight models are good enough - and I'd argue we are almost there with some of the latest releases. If you hook up the latest OpenAI models to something like OpenCode, its a taste of what an open harness with a powerful model (outside of a providers ecosystem) will be able to offer developers in the future.

Given your attitude, this comment is probably futile, but here goes nothing.

Your attitude here, to give a somewhat more illustrative automotive example, is akin to shunning many modern safety devices, standards and common sense. Driving on bare tyres is fine pretty most the time when sunny, until the road is wet, upon which you will likely end up in a ditch. Same deal with seatbelts, where you're fine for >99% of the time, until your knees end up sandwiched in the windscreen after an accident. Not to mention ABS, AEB, and a whole slew of other safety advancements.

You can keep driving your '70s wagon with bench seats and no seatbelt, no one will stop you. But when your banking details are sniped or your system is subject to a cryptolocking attack and you have to deal with the subsequent inconvenience/crisis, you know why.

If you're going to adopt "but it hasn't happened to me" attitude, you should drop the "just SHUT THE FUCK UP" attitude in your post and start ignoring those comments instead, since the people telling you to upgrade are plainly, objectively correct.

You don't need to use Windows at all to have a modern and secure computing platform, BTW. Once ads started appearing in Windows, it was clear they abandoned all reason for madness. Any power user using it as their primary OS is just asking for it at this point.

Wow; this is the first I've seen someone else describe "geometric nightmares". You're spot on in that they're "hard to describe" - they used to happen very frequently when I was younger and could never explain them.

Bicarbonate Soda powder in the litter box (dumped in and shaken around) entirely eliminated any smell of urine, at least with our two bengals. We started looking into odour treatments, we were tipped off by another savvy shopper at the pet store. Became part of our litter box routine very early on. My backyard science hypothesis is that it reacts with the odour-causing compounds in the urine (urea or uric acid?), which neutralises the smell.