edude03

Maybe the only downside to how fast LLMs are moving is papers come out faster than anyone (not at Google) can train and test the improvements.

I got into deep learning around when ReLU and dropout was hot and on my consumer 1080 I was able to change one or two lines of code and test the improvements in a few hours, whereas now, I guess I'll need to wait a few weeks for mistral et al to try it out

Possibly this https://github.com/intitni/CopilotForXcode

Tesla as a company has only existed for 10 years and the current design language 5? years or so. Anyone old enough to drive would have experienced gas cars and the norms there, but we're also starting to see people get a Tesla as their first car.

As the population ages, maybe the current "Tesla norms" will become "car norms" and this will be a moot point.

> The cars don't support Carplay, so I had to figure out how to log into some app on a touch screen. Eventually, after much frustration, I could get bluetooth to work so it could function like a car head unit from 2007.

Coming from a place of genuine curiosity, why would you prefer CarPlay over logging in on "some app"?

I currently have two cars with CarPlay and a Tesla and I _personally_ find the Tesla infotainment system works better than CarPlay (probably because of the integration).

On the other hand, I can imagine on a rental car it'd be easier to just plug your phone in via USB and have all your apps & whatever on your phone ready to go. I guess CarPlay is better if you switch cars a lot and (potentially?) the Tesla system is better if you have the same car all/most of the time?

For a counter anecdote - my tesla went in for some minor service the other day and they gave me a loaner, the loaner worked exactly like my personal car, logged into all my apps and whatever, seat, mirror temperature, drive mode etc all synchronized. I wonder if more people owned Teslas as their personal vehicles switching to another Tesla would feel more normal?

~20 years to directly answer your question.

I said this in a child comment - but to put it a different way - Imagine google didn't exist, how would you go about discovering content? When I first got on the internet there used to be directories, which was literally just a list of links sorted by topic. Although there was no "algorithm" I imagine (based on my own usage) that items listed closer to the top were more often clicked and thus grew faster than items lower on the list.

In fact, even before the internet, this phenomenon existed in the Yellow Pages which is why you have things like "aaa local plumbers" so they'd appear before their competitors.

So, I agree with you, but as a counter anecdote - I was subscribed to Nebula for ~2 years, and I barely used it because there was no algorithm telling me what things it thought I'd like. Even when I had a topic in mind that I was interested in (mega projects as a random example) I'd get low 100s of results, many from creators I had never heard of, and the ones I chose to watch would typically be from creators I knew from YouTube which of course, were presented to me by "the algorithm".

While I agree it's harmful to the web to be essentially under control of a very small number of mega corporations, even when you actively avoid them like I did in my nebula example, it's not clear what the solution to content discovery is (and of course, google et al don't really stop you from consuming content you already know about which is my problem with the original article)

I'll start by saying, I'm sorry this is happening to the retrododo team, I hadn't heard of them before this article (I'm not really intro retro games) but it seems like they make genuinely good and useful content for the community.

On the other hand, more and more these days I see articles & videos that I (possibly unfairly) summarize as "My content deserves to be prioritized by 'the algorithm' and $BIG_CORP is against me".

I'm not a full time content creator so again maybe that factors into the mentality - but I honestly don't understand why so many people seem to believe that their content not doing well by some arbitrary standard means some force is against them. To me it seems more like building your brand organically, publishing via "open" platforms (and yes, I'm aware that's getting harder and harder) and encouraging your supporters to interact with you on platforms you control would be much more sustainable than expecting 'the algorithm(s)' will provide you with your expected growth.

I don't even use google, so if I were interested in getting "the best arcade cabinet" as one of the examples the author used - I would actually be looking to either reddit, or gaming YouTubers or gaming sites I already use, which are the places I would expect to hear about Retro Dodo

9th gen intel? According to geekbench 6[0] the m1 is between 2x-5x faster with better battery life, so depending on your workload, it's quite possibly not close enough.

[0]: https://browser.geekbench.com/mac-benchmarks/

I agree with the premise that ebooks can be a better experience than paper books, I don't agree that pirate libraries can replace, or even serve a similar purpose to physical libraries though.

Can you provide some examples then?

Pretty high if you want to have storage separate from your compute (for example, editing video on one computer while storing the raw footage on a NAS, which is one of my use cases).

A single average consumer SSD can already deliver 30gbit/s, and having a few "prosumer" drives like the WD Black SN850 in raid can deliver 100gbit and cost under a hundred each.

(I'm not part of the Gitea community, just a casual user of gitea since the gogs days, so please take this for what it's worth -) I don't understand why Forges (and what feels like a large part of the free software community) is against corporate sponsorship.

The stated goal of '"ensuring the long term success of the project" by allowing companies "unable to contribute back to open source via sponsorship or code contribution" to support Gitea in the form of paying Gitea Ltd directly ' seems extremely reasonable to a passerby and didn't seem to come with any strings attached, so why the knee jerk reaction of forking the project?

Yeah, not having a way to share keys between thread networks I think is the biggest technical challenge for the user/user experience. Home Assistant for example is able to import keys from an iOS device using its app and join a HomeKit thread network, but I don't know if that approach is scalable.

> Cheap Zwave/Zigbee integrated HA hub

The ikea Dirigera and smartthings hub have great HA support from what I understand, I'm not sure what'd it take to get the HA logo slapped on the box and have them officially supported though

From what I understand the benefit of ip addressing for IoT things is that apps on your phone (for example) can communicate with them directly, ignoring the underlying transport (and therefore, no need to have a thread radio in your phone for the manufacturer specific app to work)

I've always found that the most reliable combination is the Manufacturers app + hub + IoT devices (for example all Aqara or all Nest or all Hue) and interop hasn't worked great because there are special features that only work with the app made for the device (which again, only works with that manufacturers hub) so thread should fix that by making the hub just a dumb Zigbee to IPv6 adapter.

(I've been running home automation stuff for ~10 years now but) I am little surprised the author thought wifi/bluetooth would be better than technology made specifically for (home?) automation. To me meshing is a natural fit for interconnecting devices all around the house vs the hub and spoke model that wifi has.

Furthermore, I (read) and kind of disagree that thread/matter is a mess, but I'm coming from the point of view of an engineer who is keenly interested in home automation tech; for the typical consumer that's probably a fair take.

For what it’s worth blu rays copy protection system was designed so that keys can be revoked so without ripping a DRM free version it’s also possible to lose access to your physical media in this way

_actively_ watch? Probably not many. Having it on as background noise however? 5 hours is pretty easy

Definitely. I don't think you're wrong to want that but I think people who wants teslas buy them for the tech and cloud features.

I read the comments before I watched the video and I'm actually way more optimistic about this now than I was expecting to be.

Yes, google or apple will probably ship LLMs on their phones eventually and make an overall more compelling product than Rabbit _if_ you're only willing to have one device.

Considering the relatively low price (and assuming they actually deliver it in a timely fashion) I can see this being useful as a pocket assistant between now and whenever google et al eventually catch up, and potentially past then if rabbit keeps innovating.

> A lot of Windows features depend on Hyper-V, once enabled Windows is not booted directly any more, Hyper-V is started and the main Windows system runs in a privileged VM.

Got a source for this? Not that I don't believe you but other than for the Xbox I haven't seen/can't find any details about this.

I think the intent is really different.

For LLMs you're essentially teaching them language by showing them lots of examples of written language - newspapers are of course a great example of written language.

The goal of OpenAI is not to reproduce newspaper articles verbatim when asked questions (even if the answer could be a newspaper article) and the fact that it can happen is a side effect of how LLMs work.

When a HN participant shares a (pay walled) link to a NYT article, I do want to read the exact article linked verbatim because while the facts of the article may be reproduced elsewhere in a form that's free, specific word choices or whatever might be a focal point of the discussion on HN, and therefore I can't realistically participate in a discussion without having read the article being discussed.

And as an aside, I have no problem with paying to read news, or whatever media, however it's impractical for me to subscribe to every news source HN participants link to, and therefore I gravitate to archiving services instead. I do wish there was a better solution - for example Blendle with more sources.

Go the same speed as traffic? If that turns out to be above the speed limit "they can't (and realistically won't) stop us all"

The demo apps that microsoft shows probably are what they're using it for. If you look at the microsoft store for/on the HoloLens you'll see a bunch of training and communication apps.

> Open to me means I can install any x86-64 desktop OS on it

But you claim you would have bought an arm64 machine if it could run boot camp?

Between that and wanting CUDA it sounds like you would have never bought a Mac after 2012 (the last year you could get an nvidia GPU)

It's a poor moral practice perhaps but arguably a good business practice to spend as little as possible on "cost centres"

Right, I think I get it now, like benpacker said in https://news.ycombinator.com/item?id=38685760 it's a way to map the kubernetes API to the fly platform (basically). Which makes a lot of sense!

Not sure what the implications of that are in practice but sounds interesting.

I'm confused about what this is actually offering (also very tired due to some flight problems; anyway)

To me, I'd imagine kubernetes on fly as running kind (kubernetes in docker) with fly converting the docker images to firecracker images OR "normal" kubernetes api server running on one machine then using CAPI/or a homegrown thing for spinning up additional nodes as needed.

So, what's the deal here? Why k3s + a virtual kublet?

> But there is no equivalent system for whole clusters. To transition from one cluster to another, you have to handle everything that k8s gives you without relying on k8s for it

Ah, but there are. Maybe not a cluster orchestrator that would let you hot swap a single cluster, but if you're only running one cluster it's inherently a pet already.

One common way that I've personally used to solve this problem is use Federated Clusters (kubefed) to run multiple clusters with global load balancing so that even when one entire cluster goes away its traffic is just shifted to the other cluster (whether because of upgrades or unexpected outages of the control plane)

Maybe more importantly, you could get a distribution to support you but what about upstream projects? It'd be a big lift (if not impossible) to get projects like cert-manager cilium whatever to adopt the longer release cycle as well.

Is it normal for a distribution to also package upstream projects that customers want?

I know it rarely happens in practice, but I disagree than Kubernetes needs an LTS since the Kubernetes cluster itself should be a "cattle not a pet" and thus, you should "just" spin up a new cluster with the same version as your upgrade strategy.