Hacker Newsnew | past | comments | ask | show | jobs | submit | gclawes's commentslogin

Can you hear me?

Could this be a road to getting GrapheneOS approved under Play Integrity (for contactless payments, etc)?


I don't think so, since Play Integrity is incompatible with open source.


Right, my understanding is that the whole point of "device integrity" is for a device manufacturer that your bank trusts to be able to vouch that your device is "secure". If the device is under your control rather than the manufacturer's (as would generally be the case with an open source OS), they can't credibly make such guarantees.


Maybe, probably not.


The government should be entitled to any lawful use of a product they purchase, not uses dictated solely by the provider. It's up to courts to decide what lawful use is, it's not up to these companies to dictate.


The product is a service, and they agreed to a contract. Now they don't like the contract.

Is your view that contracts with the government should be meaningless? That the government should be able to unilaterally, and without recourse, change any contract they previously agreed to for any reason, and the vendor should be forced at gunpoint to comply?

If you do believe this, then what do you believe the second order effects will be when contracts with the government have no meaning? How will vendors to the government respond? Will this ultimately help or hinder the American government's efficacy?


Seriously.

Hegseth trying to play “I’m altering the deal. Pray I don’t alter it any further” just shows this gang’s total lack of comprehension of second-order effects.


> It's up to courts to decide what lawful use is

No, it’s up to the government to create policy and legislation that outlines what is lawful or not and install mechanisms to monitor and regulate usage.

The fact that an arm of the government wants to go YOLO mode is merely a symptom of the deeper problem that this government is currently not effectual.


Do you have any insight that what they want to do is YOLO, as opposed something your sure you’ll disagree with?


YOLO here refers to unsafe usage of LLMs. Your government is supposed to make legislation that protects all of its citizens, it’s not “what you agree with” game.


Yeah, I knew what was meant. Unsafe being a moving definition by an arbitrary set of people.


You're joking, right?


Terms of Service would like to have a word....

Not like limiting uses of products is anything new


Providers are free who they choose to do business with, or not do business with. Are you arguing that the government should be able to compel a provider to allow their use when it’s well documented the government does not respect nor adhere to the rule of law? I think you misunderstand commerce and contract law.


Providers are bound by plenty of laws that alter how they do business or who they do business with.

You can’t say “no disabled people at your business”. Hell, you can’t even say “no fake service animals at my restaurant”. Many in America also think you can’t say no girls in the Boy Scouts, or no men in a women’s locker room.


When Congress makes the law, you will be accurate. At this time, there is no law that enables the US executive branch to achieve their desired outcome of strong arming Anthropic.

> Many in America also think you can’t say no girls in the Boy Scouts, or no men in a women’s locker room.

Your average American is low functioning, low education, vibe driven with a 6th-8th grade reading level, so this ("What Americans think") is not terribly relevant in my opinion. Provide statute and case law.


Defense Production Act.


Which I'm sure will work as well as tariffs. Three years to go, good luck to them, slowing them down as proven exceptionally effective. Their efforts will likely die at midterms.

https://www.justsecurity.org/107087/tracker-litigation-legal...


Anthropic says it will challenge Pentagon supply chain risk designation in court -https://news.ycombinator.com/item?id=47189441 - February 2026


Not really. Services are provided on terms acceptable to both parties. This isn't about what's legal, it's about the terms of the service agreement.


Strange take


> This contradictory messaging puts to rest any doubt that this is a strong arm by the governemnt to allow any use.

Why the hell should companies get to dictate on their own to the government how their product is used?


Every company is free to determine its terms of use. If USG doesn’t like them they should sign a contract with someone else.


Every company is free to state their terms of use, but not all have been upheld when challenged


What’s your angle here? I’m genuinely curious. If the government told you that you had to muck out portable bathrooms with your bare hands even if you didn’t want to, wouldn’t you find that objectionable?


I’m sure they would find it objectionable, just like how many reacted negatively to the draft, but it was imposed anyways.

The government should have far less control and power over individuals and businesses than it currently does.


Well, the rates are different from country to country, but everyone knows taxes. I really don't want to give away almost 40% of my income... Does anyone care what I want or like?


Taxes aren't forced labor or indentured servitude, and aren't prohibited in any democracies. They're imposed by law through the actions of our duly elected representatives.


What grounds for challenge do you imagine here?


> Every company *

* excludes tiktok


Can I run a business and say “No use by insert race here”? If they don’t like it, they can shop somewhere else, right?


Of course not, nor can you write a contract that places your customers in indentured servitude. Those would be illegal contractual terms.

But this is irrelevant to the case we are discussing, where Anthropic used legal contractual terms, and the government willingly signed them, then demanded they be changed after the fact.


Ofcourse we're gonna compare being against the use of technology for Mass surveillance/Autonomous weapons with being racist, like wtf kind of argument is this? So because businesses can't implement racist policies they shouldn't be allowed to have any policies concerning the use of their tech? Mindblowing.


Well, the question is the fine line between racism and discrimination. Or, whats the difference between misogyny and pacifism? What am I allowed to dislike? Is it already across the line if I dont like dogs? What if I had really bad experiences with dogs in the past? Is it OK now, or still not? What if my childhood was basically a crazy mess because of my mother? Am I allowed to be careful around women now? Or am I creepy because of that? What if I escaped a warzone during my childhood? Is militant pacifism OK now? What if the military saved my family from being killed? Is it OK if I am pro military budget, or am I a system-whore now?


[flagged]


If your argument is “every company is free to determine its terms of use”, except when told otherwise by the government, you’ve proven my point. The government is saying they need to provide unfettered access.


“Told” is different than it being written into law. Go update the laws first and then you have a valid argument


So they'll be able to use the already-written DPA, right?


They can try, but:

1) it’s pretty transparently obvious that Anthropic is not a supply chain risk, and that this is a retaliatory gesture. So I don’t support that usage.

2) if they do try, Congress or SCOTUS could well reduce or remove that authority. I give the Trump admin enough credit to assume that they are considering carefully which laws they spend in this way, DPA is a valuable chip they may need to spend for something more valuable than Hegseth’s temper tantrum.


Because technology companies know more about their product's capabilities and limitations than a former Fox News host? And because they know there's a risk of mass civilian casualties if you put an LLM in control of the world's most expensive military equipment?


Because the government is here to serve us. Not the other way around.


The government has a responsibility to protect its constituents. Sometimes that requires collaboration. This isn’t hard.


Is this one of those times? Seems pretty clear it's not.

The third amendment is there for a reason. I am a third amendment absolutist and willing to put my life on the line to defend it.


I wonder what you can't justify this way.


That’s a good question. Assuming a righteous and just government:

The government couldn’t justify the killing of innocent civilians.

The government couldn’t justify the killing of the unborn.

The government couldn’t justify eugenics.

There are objective moral absolutes.


Wow, that's just so many assertions and none of them follow from the statement that the government can break the law in order to protect its citizens. In all of those cases I can just say "they can if it is to protect its citizens". Remember, the premise here is that you are performing the act in order to protect constituents. So before all of those statements you have to assume "They are doing this in the genuine believe that it protects constituents".

The argument so far seems to be "They can do anything, but there are moral absolutes that I can personally list out, and in those cases they can't do those things". That is a hilariously stupid view of the world but sadly a common one.

Even if I grant moral objectivity, I reject that you have epistemic access to it so it's moot.


I normally don't respond to bad faith responses like this, but I found the following quote pretty funny:

> Even if I grant moral objectivity, I reject that you have epistemic access to it so it's moot.

This is a silly and self refuting statement.


> This is a silly and self refuting statement.

No it isn't and it's a pretty standard argument.

Other than insulting you, my response was pretty damn charitable tbh. I tried to state your argument for you as best I could.


Same reason they cant quarter troops in your house: the law



> Why the hell should companies get to dictate on their own to the government how their product is used?

Well:

"""

Imagine that you created an LLC, and that you are the sole owner and employee.

One day your LLC receives a letter from the government that says, "here is a contract to go mine heavy rare earth elements in Alaska." You don't want to do that, so you reply, "no thanks!"

There is no retaliation. Everything is fine. You declined the terms of a contract. You live in a civilized capitalist republic. We figured this stuff out centuries ago, and today we have bigger fish to fry.

"""

* https://x.com/deanwball/status/2027143691241197638


This is a terrible analogy. Imagine you’re an LLC that signed a contract to mine minerals, but your terms state you’d only mine in areas you felt safe. OSHA says it’s safe but you disagree, because….. any number of reason unknowable to an outsider. Maybe you just don’t like this OSHA leadership. That is more like what is happening.

Signing a contract with Anthropic assuming they wouldn’t rug pull over their own moral soapbox was mistake number one.

I love anthropic products and heavily use them daily, but they need to get off their high horse. They complain they’re being robbed by Chinese labs - robbed of what they stole from copyright holders. Anthropic doesn’t have the moral high ground they try to claim.


The (hypothetical) contract is clear, though. The condition is stated in objective terms: “in areas you felt safe.” If the Government agrees to this, then they should be bound just like any private counterparty would. If the Government didn’t agree to this, they should have negotiated that term out in favor of their preferred terms.


I agree. Which is why I said signing a contract with anthropic was a terrible idea in the first place.


Is it a rug pull? Where in the terms of service does anthropic say their models can be used for autonomous weapons and mass domestic surveillance?


What's the benefit of this over tinygrad?


Completely different layer; tinygrad is a library for performing specific math ops (tensor, nn), this is a compiler for general CUDA C code.

If your needs can be expressed as tensor operations or neural network stuff that tinygrad supports, might as well use that (or one of the ten billion other higher order tensor libs).


Don't Windows Hello camera devices have some kind of hardware attestation? I'm sure verification schemes like this will eventually go down that path soon.

My guess is that's probably one of the reasons Google tried to push for Play Store only apps, provide a measurable/verifiable software chain for stuff like this.


That the camera is real doesn't imply the thing it's viewing is real.


You're not wrong, but I have had to do video verification over a phone once, and it seemed quite advanced. It would flash through a number of colors and settings and take probably 30 frames of you. I presume they're checking for "this came from a screen and not a human", but of course I have no idea how it works, so I don't know if it's truly sophisticated or not.


As I understand it, 'Windows Hello' requires a near-IR image alongside the RGB image.

It's not the fancy structured light of phone-style Face ID, but it still protects against the more common ways of fooling biometrics, like holding up a photo or wearing a simple paper mask.


Fair enough. That removes the virtual option, and you'll be forced to point the camera at your older brother.


Windows Hello cameras are all "depth" cameras so a flat photo won't pass muster.


Two flat images, one for each of the sensor's camera


That’s not how they work. They emit structured light in the form of an array of infrared dots and they measure the time of flight to where the dots strike something.

Maybe new ones are different but that’s how they used to be. Little Kinect devices, really, for sensing faces instead of whole people.


You are exactly right. There's a description here:

https://learn.microsoft.com/en-us/windows-hardware/design/de...

These cameras are considered a "secure biometric" device and AFAIK nobody has faked them. I've flagged the poster who said "try two flat images"


You don't need to flag people who make a mistake.


>I've flagged the poster who said "try two flat images"

Hello there! It appears you are misapplying the flagging system. While the suggestion may be incorrect, it is not an "egregious comment".

In addition, your comment doesn't follow the Hacker News Guidelines:

https://news.ycombinator.com/newsguidelines.html

Don't feed egregious comments by replying; flag them instead. If you flag, please don't also comment that you did.

Have a great day!


Did you report your neighbours for skirting covid rules to?


Yes they do. Part of the reason why you can't use certain webcams that are Windows Hello compatible (I.e. with IR) in recent versions of Windows.


IIRC they had these fingerprinting pads?


If I could use the Oxide stack in a homelab form factor, I would be so happy...


You can! There are plenty of us running various minis or old equipment. The non-gimlet deploy pattern supports virtual networking with x86_64 "sleds."


Can you clarify what parts of the Oxide stack you're running? Is it the entire control plane?


I was thinking this, too. Here's an even crazier thought: don't even make it rack mount. Make it NUC-sized. Two PB&Js stacked on top of each other, that's the form factor. EC2 except it lives under the couch.


Their engineered power and cooling solutions are all for rack-scaled hardware, doing a NUC wouldn't make sense. Now a silent and efficiently cooled studio-sized rack with enough hardware (including AMD GPUs) for reasonably quick AI inference with the latest local models, that's something that they could do and be quite popular.


Framework desktop mainboards 128GB version. They even show a mini rack with 4 of them.


That's mini ITX hardware, not really rack-sized in the sense of a common studio rack.


Right, it's the perfect size for a 10" mini rack designed primarily for home labs. Which is the form factor shown on the framework desktop mainboard product page.


10 in. wide mini racks are fine if you have very limited space at your disposal, but it's unclear that Oxide could bring much value there. The 19 in. width is not just for enterprise server equipment, it's in common use at studios that do all sorts of creative media production. That's where there's a really strong case that Oxide's rack-scaled solution (though of course in a reduced-height, custom format) could bring a lot of value, particularly given the recent explosive growth of local AI.


I'm not saying they should put their software stack on a Celeron (or whatever Intel calls their cheap CPUs now). But no racks, please. I just can't get with the "rack in my house" crowd. If you have a basement, then fine I guess. But I live in an apartment, and I don't have the space or patience for a computer that's the size of a mini fridge, and sounds like a jet engine.


All of their software is open-source, including the firmware. I bet this is actually possible for a subset of their tools.

Step 1 could be to get Illumos running on a local x86-64 machine.


I'm interested, tell me more. What about Oxide attracts a homelab user?


Many homelab users are actually building things out in an effort to learn tooling that they will then use at work. Or just out of intellectual interest.


Or to get skills to get a better job.


Yeah, a small-scale rack for home would be great to replace the beowulf cluster me and others are still stuck with. I'd probably pay a premium for it, given what I can tell from their product material.


If they can scale down the hardware to something close to homelab-ish in price, would be great marketing and way to build expertise to have their big boy solution promoted at workplaces. Probably not a priority at their stage though.


Prices start around 800k last time I heard, I don't know if that fits within what you consider a premium or not :-)


Hah, that might be slightly above what I'm ready to pay for a at-home server yes :) But given the right specifications and software integration, I'd probably be ready to pony up up to somewhere around 10K for a complete solution if it could replace all my existing hardware, even if it was more expensive than other options with worse tradeoffs.


You would need 3-phase power too. At 208V and 15kW it'll draw over 70A peak :P. If you can wire that up in your living room, I raise my glass to you!


Nothing a licensed electrician couldn't figure out, although with 230V plugs instead and probably closer to the garage than our living room :)


You have 3-phase service to your house? Lucky. I wish I could get 3-phase here but I think it would require trenching like half a mile to the main road. I get by with a phase converter on my lathe but it's significantly less efficient.


I'll take the challenge if it gets me Oxide hardware!


The rack has custom Intel Tofino based switch that uses P4 to do network magic. Hard to put this in a home lab.


That sounds like a challenge.

All I need is some financial backers and I'll happily document the whole thing on a blog and a youtube channel. Now accepting offers.


Starlink satellites also radiate a non-trivial amount of the energy they consume from their phased arrays



I very much want a Framework 13 with a decent AI chip, but the abysmal battery life is a hard stop for me, sadly. I know this isn't entirely on Framework (get your shit together AMD).


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: