lifis

You can just fork the repository, give it access to the fork and then merge what you want

I think it's ok to be signed-in when opening your own repositories, but definitely not when opening repositories from other accounts. And also the webview keyboard shortcut thing needs to be fixed to only allow harmless keybinds and NOT propagate to any keydown handler. Also on desktop it should be removed in favor of Electron intercepting directly. And on web it should probably disabled by the default.

Huh? It's not advertising, it's disclosure that the code was not fully (or at all) written by you.

The expected purpose of websites is to spread information, so whether users get it by making a request to your website or to Google is irrelevant. In fact, if they get it from Google it's better because it reduces website load.

If instead the purpose of your website is to manipulate users for financial gain (for instance by showing media attempting to manipulate their purchasing decisions, after receiving a bribe from a vendor), and the information is just a way to lure users, then maybe this malicious business model will finally be no longer possible.

If they want a dependently typed language, why not use one? Lean is good, and I don't think it has any significant downside wrt Haskell other than more limited library ecosystem (but I guess AI can translate Haskell libraries to Lean very effectively).

AI is not yet at the superhuman stage where you can tell it "clone Photoshop" and get a perfect result within a day for almost free

Seems bullshit, apparently it only works with TPM-only mode, which is obviously insecure (it relies on neither the OS nor the hardware being exploitable, on a random Windows PC...), and not worth building a backdoor for.

The way one would backdoor something like Bitlocker is to encrypt the disk encryption key with a (post-quantum) public key for which only the backdoor owner has the private key for, and then put it on a place on disk that is unused by the filesystem.

Surely one can just escape the input, no? Seems astonishing if someone isn't doing that

That is only the case if people enter exclusive relationships. But if someone has access to a dating app or system that works really well, there is little reason to do that.

Not clear what it actually does, but seems equivalent to a global right click menu with "Chat with AI about this"

This seems easily solved with a tool use hook that calls git add .; git commit a -m "<tool description>", specifying an alternate .git directory if desired

Why? From searches and LLMs it seems it costs $50-100 to move a tonne 1000 km via truck, giving 0.05-0.10 $/kg for a supermarket 500km away. Fruit prices at at least $4.5/kg for peaches, 3.75$/kg for apples 1.45$/kg. So transport cost seems negligible and if fruit is given away for free, it seems it would be very profitable for any supermarket in region to show up with a truck. What's missing in this analysis?

It's equivalent to multiplying all inputs by log b. And multiplying all inputs by a value changes how much the probabilities are extremized. This is easy to see because adding a value to everything doesn't change the output, so the biggest input can be assumed to be 0 and others negative. So multiplying by 0 makes all outputs equal while as the multiplier tends to infinity, all other inputs tend to -infinity and thus the biggest output tends to 1 and others to 0. Multiplying by negative numbers results in the lowest becoming the highest.

The Linux kernel is not usable as a security boundary, so anyone who wants to do "shared hosting" and not be hacked needs to use something else, like gVisor or firecracker VMs

The only important system that uses it as a security boundary is Android and there is mitigated by the fact that APKs need user approval, plus strict SELinux and seccomp policy plus the GrapheneOS hardening, and in this case the mitigations succeeded (https://discuss.grapheneos.org/d/35110-grapheneos-is-protect...)

Well, that was quite predictable.

Absurdly high price for a novel device of unclear utility (a VR headset but incompatible with all existing VR software) resulting in few users.

No support for PC VR nor Android/Quest VR apps resulting in little software, no massive investment in getting Vision Pro specific software written, little interest in porting due to the few users.

And what do you think society/culture is?

It's a set of biases installed in people, whose purpose is mostly to replicate themselves.

Humans are MORE susceptible that LLMs, because LLMs's biases are easily steered to something else, unlike most humans.

I think you can fix this by either patching the binary and replacing the offending prompt with an empty string, or by pointing the harness to an API proxy that filters it out

You don't seem to realize that humans also work this way.

If you ask a human why they did something, the answer is a guess, just like it is for an LLM.

That's because obviously there is no relationship between the mechanisms that do something and the ones that produce an explanation (in both humans and LLMs).

An example of evidence from Wikipedia, "split brain" article:

The same effect occurs for visual pairs and reasoning. For example, a patient with split brain is shown a picture of a chicken foot and a snowy field in separate visual fields and asked to choose from a list of words the best association with the pictures. The patient would choose a chicken to associate with the chicken foot and a shovel to associate with the snow; however, when asked to reason why the patient chose the shovel, the response would relate to the chicken (e.g. "the shovel is for cleaning out the chicken coop").[4]

On LMArena, Claude Opus is ranked as the best at everything except image and video generation, which it does not support. That may be inaccurate, but it's plausible

I'm really surprised that:

1. Anthropic has not published anything about why they made the change and how exactly they changed it

2. Nobody has reverse engineered it. It seems easy to do so using the free token counting APIs (the Google Vertex AI token count endpoint seems to support 2000 req/min = ~3million req/day, seems enough to reverse engineer it)

Obviously an image picker shouldn't leak filenames... The filename is a property of the directory entry storing the file storing the image. The image picker only grants access to the image, not to directories, directory entries or files.

If you want filenames, you need to request access to a directory, not to an image

The paper somehow seems to be missing the most interesting part, i.e. the optimal constructions of functions from eml in a readable format.

Here is my attempt. I think they should be optimal up to around 15 eml.nodrs, the latter might not be:

# 0

1=1

# 1

exp(x)=eml(x,1)

e-ln(x)=eml(1,x)

e=exp(1)

# 2

e-x=e-ln(exp(x))

# 3

0=e-e

ln(x)=e-(e-ln(x))

exp(x)-exp(y)=eml(x,exp(exp(y)))

# 4

id(x)=e-(e-x)

inf=e-ln(0)

x-ln(y)=eml(ln(x),y)

# 5

x-y=x-ln(exp(y))

-inf=e-ln(inf)

# 6

-ln(x)=eml(-inf,x)

ln(ln(x))=ln(ln(x))

# 7

-x=-ln(exp(x))

-1=-1

x^-1=exp(-ln(x))

ln(x)+ln(y)=e-((e-ln(x))-ln(y))

ln(x)-ln(y)=ln(x)-ln(y) # using x - ln(y)

# 8

xy=exp(ln(x)+ln(y))

x/y=exp(ln(x)-ln(y))

# 9

x + y = ln(exp(x))+ln(exp(y))

2 = 1+1

# 10

ipi = ln(-1)

# 13

-ipi=-ln(-1)

x^y = exp(ln(x)y)

# 16

1/2 = 2^-1

# 17

x/2 = x/2

x2 = x2

# 20

ln(sqrt(x)) = ln(x)/2

# 21

sqrt(x) = exp(ln(sqrt(x)))

# 25

sqrt(xy) = exp((ln(x)+ln(y))/2)

# 27

ln(i)=ln(sqrt(-1))

# 28

i = sqrt(-1)

-pi^2 = (ipi)(ipi)

# 31

pi^2 = (ipi)(-ipi)

# 37

exp(xi)=exp(xi)

# 44

exp(-xi)=exp(-(xi))

# 46

pi = (ipi)/i

# 90+x?

2cos(x)=exp(xi)+exp(-xi))

# 107+x?

cos(x) = (2cos(x))/2

# 118+x?

2sin(x)=(exp(x*i)-exp(-xi))/i # using exp(x)-exp(y)

# 145+x?

sin(x) = (2sin(x))/2

# 217+3x?

tan(x) = 2sin(x)/(2cos(x))

They have ACP servers but they might apply different rate limits or policies if they notice ACP use, while a solution like yours would not trigger that unless it becomes popular enough to specifically detect. It also seems this provides more features them just an ACP server.

That's cool but by only supporting Claude Code you are contributing to the Anthropic lock-in problem.

This needs to support at least Gemini CLI, Codex and OpenCode as well, preferably by being generic as much as possible.

I can't understand why address instability is a problem: if a Mutex is moved, then it can't be locked (because you need to hold a borrow while locked, which impedes moving), so using addresses is perfectly fine and there is absolutely no need to use IDs.

Also the fact that it doesn't detect locking the same mutex twice makes no sense: a static order obviously detects that and when locking multiple mutexes at the same level all you need to do is check for equal consecutive addresses after sorting, which is trivial.

Overall it seems like the authors are weirdly both quite competent and very incompetent. This is typical of LLMs, but it doesn't seem ZlLM-made.

But why doesn't the CPU just lock two cachelines? Seems relatively easy to do in microcode, no? Just sort by physical address with a conditional swap and then run the "lock one cacheline algorithm" twice, no?

Perhaps the issue it that each core has a locked cacheline entry for each other core, but even then given the size of current CPUs doubling it shouldn't be that significant. And one could also add just a single extra entry and then have a global lock but that only locks the ability to lock a second cacheline.

Vanadium, Chrome and Firefox (all for Android) all accept all the revoked certificates... But revoked.badssl.com is considered revoked

I don't think you can steal Bitcoin with a quantum computer because the blockchain only stores the 256-bit hash of the public key, so you need to reverse that, which costs 2^128 with grover's algorithm

Are you disabling them for paying subscribers?

I don't understand this: isn't the thing in the article only relevant for software simulations, while in hardware ordering is arbitrary like in Verilog, or at least dependent on wire lengths that are not specified in HDL? (unless you delay the effect to the next clock update, which it seems to me will work the same in all HDLs and targets).

And afaik HDLs are almost exclusively used for hardware synthesis, never seen any software written in those languages.

So it doesn't seem important at all. In fact, for software simulation of hardware you'd want the simulation to randomly choose anything possible in hardware, so the Verilog approach seems correct.