HNHacker News
TopNewBestAskShowJobs

coppsilgold

1,156 karma · joined February 24, 2023

email: coppsilgold@protonmail.com
submissionscomments
coppsilgold··on Google broke reCAPTCHA for de-googled Android users
As far as I know no currently proposed age verification method does this in practice.

The only way to implement truly privacy preserving age verification is through zero knowledge proofs (or blind signatures) but what that would allow is undetectable token forging.

coppsilgold··on Google broke reCAPTCHA for de-googled Android users
Realistically, what Google will do in such a scenario is collect data about the illicit service, enumerate the devices the farm uses and what other activities the devices participate in. What you suggested has far less control over the devices that generate the attestations and it will show.

Also, if the implementation is competently done the phone will show the website for which you scanned the QR code. A user would be able to see whether or not that matches the site where they observed the QR code and proceed accordingly. In time Google will probably integrate it into the Chrome browser where a proxied QR code cannot even be shown.

coppsilgold··on Google broke reCAPTCHA for de-googled Android users
Apple has their own remote attestation infrastructure and you will not be able to impersonate an Apple device without extracting private key material from the secure enclave of a legitimate Apple device or compromising Apple certificate authority private keys.
coppsilgold··on Mojo 1.0 Beta
Python is basically the master glue language at this point.

If more than a few percent of execution time is spent in Python you are probably doing it wrong.

Personally I don't even understand why Cython is a thing, just write performance critical functions in other languages:

<https://pypi.org/project/rustimport/>

<https://pypi.org/project/import-zig/>

Note that you can even start threads in those languages and use function calls as pseudo-RPC. All without an overly complex build system.

coppsilgold··on Google broke reCAPTCHA for de-googled Android users
My understanding is that this new reCAPTCHA is basically just remote attestation.

Remote attestation doesn't use blind signatures (as that would be 'farmable') so tying the device to the 'attestee' is technically possible with collusion of Google servers: EK (static burned-in private key) -> AIK (ephemeral identity key in secure enclave signed by a Google server) -> attestation (signed by AIK). As you can see if the Google server logs EK -> AIK conversions an attestation can be trivially traced to your device's EK. This is also why we don't really see and probably never will see online services which offer fake remote attestations, as it will be pretty obvious that the next step of running such a service is getting Google as a customer and having all your devices blacklisted. Private farms probably won't last long either as I'm sure Google logs everything and will correlate.

Unless something special is done with this new reCAPTCHA not only are you locking internet services behind TPM chips but you are also surrendering anonymity to Google. Unless you acquire untraceable burners for every service, the new reCAPTCHA will be technically capable to tying all your accounts across all these services together. Much like age verification. It may appear that the service would need to cooperate to link the reCAPTCHA session to your registration but the registration time alone will likely be sufficient (the anonymity set will be all but destroyed).

coppsilgold··on New research suggests people can communicate and practice skills while dreaming
Checking clocks for consistency. Text as well. They are less reliable. Some people swear by rotating a text containing object upside down and see if the text auto-rotates, apparently it does in their dreams. Some people can't read anything in their dreams.
coppsilgold··on New research suggests people can communicate and practice skills while dreaming
Training yourself to remember dreams by writing them down before they fade away is paramount, it's not enough to just think about them - they still somehow fade away along with your thoughts about them. Then read what you wrote before going to sleep again.

If you want to achieve lucid dreaming consistently you also have to develop a habit of doing reality checks. The most effective one is to pinch your nose and try to breath through it, in your dreams it will almost always work and the surprise is major.

Lucid dreaming even works for people with aphantasia: <https://en.wikipedia.org/wiki/Aphantasia>.

coppsilgold··on EU Age Control: The trojan horse for digital IDs
It seems unlikely that a true Zero Knowledge Proof system for things like age verification would ever be allowed.

Also, remote attestation doesn't work that way and for good reason. Under a true ZKP system, a single defector (extracted/leaked/etc key) would be able to generate an infinite number of false attestations without detection.

coppsilgold··on Simulacrum of Knowledge Work
> The training doesn't evaluate "is the answer true" or "is the answer useful." It's either "is the answer likely to appear in the training corpus" or "is the RLHF judge happy with the answer." We are optimising LLMs to produce output which looks like high quality output.

It's not quite as dire as this. One of the main reasons why LLM's are getting better over time is that they are used themselves to bootstrap the next generation by sifting through the training set to do 'various things' to it.

People often forget that the training corpus contains everything humanity ever produced and anything new humanity will produce will likely come from it as well. Torturing it with current generation models is among the most productive things you can do to improve the next generation systems.

coppsilgold··on There Will Be a Scientific Theory of Deep Learning
Comparing Deep Learning with neuroscience may turn out to be erroneous. They may be orthogonal.

The brain likely has more in common with Reservoir Computing (sans the actual learning algorithm) than Deep Learning.

Deep Learning relies on end to end loss optimization, something which is much more powerful than anything the brain can be doing. But the end-to-end limitation is restricting, credit assignment is a big problem.

Consider how crazy the generative diffusion models are, we generate the output in its entirety with a fixed number of steps - the complexity of the output is irrelevant. If only we could train a model to just use Photoshop directly, but we can't.

Interestingly, there are some attempts at a middle ground where a variable number of continuous variables describe an image: <https://visual-gen.github.io/semanticist/>

coppsilgold··on FBI looks into dead or missing scientists tied to NASA, Blue Origin, SpaceX
One more addition to the conspiracy theories:

    The frequency of fireballs in our planet’s skies seemed to grow in recent months. NASA and other meteor experts can’t agree on what explains it.
...

    In response to growing public interest, a NASA public affairs official said in a blog post at the end of March, “While it may seem like meteor reports and sightings have been more frequent recently, it is not out of the ordinary.” The post explained that from February to April, there is often a 10 to 30 percent increase in the number of extremely luminous meteors — and nobody is quite sure why.


    Mr. Hankey said that this 10 to 30 percent increase was already baked into the American Meteor Society tally, and that it doesn’t explain the apparent doubling of fireball sightings in the year’s first quarter.
<https://www.nytimes.com/2026/04/21/science/march-fireballs-m...>
coppsilgold··on Even 'uncensored' models can't say what they want
<https://en.wikipedia.org/wiki/Supernormal_stimulus>
coppsilgold··on Even 'uncensored' models can't say what they want
This is more or less what happens. These models are tuned with reinforcement learning from human feedback (RLHF). Humans give them feedback that this type of language is good.

The notorious "it's not X, it's Y" pattern is somewhat rare from actual humans, but it's catnip for the humans providing the feedback.

coppsilgold··on Scientific datasets are riddled with copy-paste errors
Indeed, the niches of smart cheats or smart criminals have a lot of room. Because the trajectories to reach that stage without being caught by a legal (good legal work) or moral (good person) attractor are sparse and that makes them somewhat rare.

Handwaving correlations between cheating/criminality and most personality/intelligence aspects is an error, not least because there is a selection bias problem (eg. who gets caught).

coppsilgold··on Scientific datasets are riddled with copy-paste errors
What should give people pause is how not complicated (I'd hesitate to say easy) it would be to create a python script that would generate fake data such that it would be all but impossible to determine whether it's real or not. You just need to model the measuring device and hypothesis you want to support, then sample away.

The people who get caught red handed like this are lazy, incompetent and stupid. Makes you wonder what about the ones not getting caught.

coppsilgold··on Rewriting Every Syscall in a Linux Binary at Load Time
SECCOMP_RET_USER_NOTIF seems to involve sending a struct over an fd on each syscall. Do they really use it? Performance ought to suffer.

Also gVisor (aka runsc) is a container runtime as well. And it doesn't gatekeep syscalls but chooses to re-implement them in userland.

coppsilgold··on Rewriting Every Syscall in a Linux Binary at Load Time
You mentioned SECCOMP_RET_TRACE, but there is also SECCOMP_RET_TRAP[1] which appears to perform better. There is also KVM. Both of these are options for gVisor: <https://github.com/google/gvisor>

[1] <https://github.com/google/gvisor/blob/master/pkg/sentry/plat...>

coppsilgold··on Qwen3.6-35B-A3B: Agentic coding power, now open to all
Much like a developer can insert a backdoor as a "bug" so can an LLM that was trained to do it.

One way you could probably do it is by identifying a commonly used library that can be misused in a way that would allow some kind of time-of-check to time-of-use (TOCTOU) exploit. Then you train the LLM to use the library incorrectly in this way.

coppsilgold··on GAIA – Open-source framework for building AI agents that run on local hardware
Nvidia went through a lot of effort to make CUDA operational on their entire lineup, and they did it before deep learning even took off.

You do this thing not because you expect consumers with 5 year old hardware to provide meaningful utilization but as a demo ("let me grab my old gaming machine and do some supercomputing real quick") and a signal that you intend to stay the course. AMD management hasn't realized this even after various Nvidia people said that this was exactly why they did it, at some point the absence of that signal is a signal that the AMD compute ecosystem is an unreliable investment, no?

coppsilgold··on Small models also found the vulnerabilities that Mythos found
LLMs are wordsmith oracles. A lot of effort went into trying to coax interactive intelligence from them but the truth is that you could have probably always harnessed the base models directly to do very useful things. The instruct tuned models give your harness even more degrees of freedom.

A while ago, the autoresearch[1] harness went viral, yet it's but a highly simplified version of AlphaEvolve[2][3][4].

In the cybersecury context, you can envision a clever harness that probes every function in a codebase for vulnerabilities, then bubbles the candidates up to their callsites (and probes whether the vulnerability can be triggered from there) and then all the way to an interface (such as a syscall) where a potential exploit can be manifested. And those would be the low hanging fruit, other vulnerabilities may require the interplay of multiple functions. Or race conditions.

[1] <https://github.com/karpathy/autoresearch>

[2] <https://deepmind.google/blog/alphaevolve-a-gemini-powered-co...>

[3] <https://arxiv.org/abs/2506.13131>

[4] <https://github.com/algorithmicsuperintelligence/openevolve>

coppsilgold··on Great at gaming? US air traffic control wants you to apply
If you were to design an entire ATC system from scratch (pilot interfaces, sensors everywhere in the airport and planes etc) it can be automated. But with pilots having to actually talk to ATC (and sometimes talk over each other with no feedback) instead of observing their status on a screen and pressing buttons on what they want to do or change their status it seems like it will be quite hopeless for quite some time.

What you can probably do is create software which observes traffic and simulates it into the future and notifies the human ATCs about risks. It might even be a good idea to try and digitize it for the ATCs so they talk less and press buttons more (which will feed into the simulation) and use TTS for the legacy transmissions to pilots that don't have an updated interface. Given the regulation on that industry it seems unlikely anyone competent enough to do it will have an interest to even try.

coppsilgold··on Reverse engineering Gemini's SynthID detection
> I'm curious - as new models are being built constantly and they're naturally non-deterministic, do you think it's possible for end users to prove that?

How is the model relevant? The models are proprietary and you never see any of its outputs that haven't been watermarked.

coppsilgold··on Reverse engineering Gemini's SynthID detection
Inserting an undetectable 1-bit watermark into a multi megapixel image is not particularly difficult.

If you assume competence from Google, they probably have two different watermarks. A sloppy one they offer an online oracle for and one they keep in reserve for themselves (and law enforcement requests).

Also given that it's Google we are dealing with here, they probably save every single image generated (or at least its neural hash) and tie it to your account in their database.

coppsilgold··on Who is Satoshi Nakamoto? My quest to unmask Bitcoin's creator
> Satoshi supported big blocks in his writings and empowered the pro-big block Gavin when he disappeared. Adam is a well known supporter of small blocks, ultimately the "winning" side of the debate. They are not the same person.

From the article:

    Then, out of the blue, Satoshi appeared on the list with an email that neatly dovetailed with Mr. Back’s position. It was the first time Satoshi had been heard from in more than four years, other than a five-word post the previous year denying a Newsweek article’s claim to have unmasked him.

    Many in the Bitcoin community questioned the new email’s authenticity since another of Satoshi’s email accounts had been hacked. But Mr. Back argued that the email sounded real. In a series of tweets, he called Satoshi’s observations “spot on” and “consistent with Satoshi views IMO” and took to quoting from the email.

    Mr. Back was likely correct: To this day, there is no evidence to indicate the email was a forgery, and no other emails from that account have surfaced.

    The Satoshi email sounded a lot like Mr. Back had in his posts during the preceding weeks, although no one took notice. Like Mr. Back, Satoshi argued that the Bitcoin network’s increasing centralization jeopardized its security. He called the big block proposal very “dangerous” — the same term Mr. Back had used repeatedly. He also used other words and phrases Mr. Back had used: “widespread consensus,” “consensus rules,” “technical,” “trivial” and “robust.”

    At the end of the email, Satoshi denounced Mr. Andresen and Mr. Hearn as two reckless developers trying to hijack Bitcoin with populist tactics and added: “This present situation has been very disappointing to watch unfold.”
It also happened to be densely cited with hyperlinks:

https://gnusha.org/pi/bitcoindev/6EC9DDF352DC4838AE9B088AB37...

https://x.com/adam3us/status/632928398893907968

https://x.com/adam3us/status/632650884011458560

https://x.com/adam3us/status/632923680104841220

https://x.com/adam3us/status/632919411112849410

https://gnusha.org/pi/bitcoindev/CALqxMTHfU5+1ezP-Jnn5obpd62...

https://gnusha.org/pi/bitcoindev/CALqxMTGBt7MNs5YWf8QzKe+4Fr...

https://gnusha.org/pi/bitcoindev/CALqxMTFC7zBN9GvHAZLQj4SbXj...

https://gnusha.org/pi/bitcoindev/CALqxMTFu6DRVMSLsGDa6AgVX1X...

https://gnusha.org/pi/bitcoindev/CALqxMTG7+MMN50VH9-Y++B1_De...

https://gnusha.org/pi/bitcoindev/CALqxMTGCkTZAs74bXk57L6JWK2...

https://gnusha.org/pi/bitcoindev/CALqxMTH_5rtOs=aSNiVrfsG_sq...

coppsilgold··on Who is Satoshi Nakamoto? My quest to unmask Bitcoin's creator
The author has collected more than enough entropy to single out Mr. Back, especially when the anonymity set of who could be Satoshi is so small.

It's either Back or someone who tried to frame him, long before Bitcoin was even remotely successful. Generally, framing someone like this is a poor strategy because it places you in the person's radius as opposed to being absolutely anyone.

coppsilgold··on System Card: Claude Mythos Preview [pdf]
It only makes sense to rent out tokens if you aren't able to get more value from them yourself.

I would go a step further and posit that when things appear close Nvidia will stop selling chips (while appearing to continue by selling a trickle). And Google will similarly stop renting out TPUs. Both signals may be muddled by private chip production numbers.

coppsilgold··on System Card: Claude Mythos Preview [pdf]
Likely an improvement on:

> We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.

<https://arxiv.org/abs/2502.05171>

coppsilgold··on From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem
There are also interesting approaches to more directly compress a large document or an entire codebase into a smaller set of tokens without getting the LLM to wing it. For example, Cartridges: <https://hazyresearch.stanford.edu/blog/2025-06-08-cartridges>

They basically get gradient descent to optimize the KV cache while freezing the network.

coppsilgold··on AMD's Ryzen 9 9950X3D2 Dual Edition crams 208MB of cache into a single chip
There may be server workloads for which the L3 cache is sufficient, would be interesting if it made sense to create boards for just the CPU and no memory at scale.

I imagine for such a workload you can always solder a small memory chip to avoid having to waste L3 on unused memory and a non-standard booting process so probably not.

coppsilgold··on Olympic Committee bars transgender athletes from women’s events
> Fun fact: "open divisions" only last as long as men are winning them. Women often outshoot men, and after Shan Zhang's win they were siloed into their own division.

That decision was made before her win.

> the International Shooting Union, at a meeting in April of 1992, and therefore ahead of the Games, elected to bar women from shooting against men in future events.

<https://www.nytimes.com/athletic/2753773/2021/08/05/in-tokyo...>

← PreviousPage 2 of 11Next →