patrakov

You could have also used "fridump" [0] to dump the app memory and search for strings that look like TOTP secrets.

[0] https://github.com/rootbsd/fridump3

The article, if examined by a non-specialist like me, seems to contain an answer to this concern at the end:

> There are papers that claim to have factored 21 with a quantum computer. For example, here’s one from 2021 [1]. But, as far as I know, all such experiments are guilty of using optimizations that imply the code generating the circuit had access to information equivalent to knowing the factors.

[1] https://arxiv.org/abs/2103.13855

Important: this is a checklist for privacy, not for general security.

Be careful with this statement. The whole premise behind banks requiring non-rooted phones is "we can be sure that sandboxing works on the original ROMs—e.g., it will prevent malware from screenshotting our app, and we know that certain custom ROMs patch this snapshot-prohibition code out, thus deliberately breaking the sandbox that we rely upon".

The elephant in the room is not addressed: software in Google Play with so many antifeatures that it can only be called malware (except that Google doesn't call it malware because it brings revenue), and no alternatives except apps outside Google Play that are not signed by a developer who would submit their identity to Google.

It does not matter whether it works. It is a contract violation and, for some banks, a reason to debank you if found - check the bank app's usage terms and conditions.

ZeroSSL - they provide Sectigo certificates under the hood. Works well with Dehydrated.

I have a horror thought: "We cannot validate your identity as you are of the wrong nationality; therefore, you are not allowed to publish any Android apps."

You can still bring a foreign SIM card. 100% effective (via data roaming) at bypassing the firewall, but expensive.

It is not sufficient to disable backspace. This thing fails the authenticity test as follows:

    * Type a few letters
    * Use the left arrow several times to go to the beginning of the line
    * Type something else

Expected result: overlapping letters, or, if the same letter has been typed twice, bold letters.

Actual result: new letters are inserted - something that a typewriter cannot do.

No. The verifiable part receives an already-decrypted copy of your traffic and mixes it with everyone else's traffic. Source: https://vp.net/l/en-US/technical#:~:text=cryptographic%20dat...

I have not inspected whether the procedure suggested for verifying the enclave contents is correct. It's irrelevant if you need to prove that the decrypted traffic, while still being associated with your identity, goes ONLY into the enclave and is not sent to, let's say, KGB via a separate channel.

It's because the buyers did not check the independent lab tests of their LED bulbs, available from https://lamptest.ru/ (sorry, available in Russian only).

The scary part is not about losing her phone. It's about having to keep the old, no-longer-secure Android phone alive just for passkeys after getting a shiny (and secure) new iPhone.

What makes this "AI will replace you with a 10x AI-based engineer" narrative a non-starter is actually having a 10x non-AI engineer on the team.

> Of course ideally there'd be two people, one doing the medical stuff and the other then documentation, but health care is expensive enough as it is.

In the 1980s USSR, every doctor actually had a nurse who did the paperwork. And somehow, healthcare was still free.

And by the way, when I was a child, even before the computers came, here is how it worked in Russia.

The doctor was listening to my breathing, looking at the throat, asking me and my mother questions, and saying various medical phrases to her assistant, who was then writing them into my patient records (a thick paper notebook).

This essay resonated with me because it highlighted the similarity between AI-written texts, describing the result as a word salad. And this also reminded me about some words from my teacher of Russian Literature: that the "bright future" themed novels of the pre-WW2 Soviet writers — works produced under strict political control — read like one big novel without a beginning and an end, and not as separate works.

And this grayness and sameness is what happens when people are forced to "think" as a chorus, either by the authorities or their censorship, or voluntarily by using the same AI's help.

> If it's a choice between filtering out people who freeze up at interviews, or letting in scammers who have no ability whatsoever, I'm afraid the first is still the best option.

So far, my opinion is: you still can weed out scammers by running your questions against an LLM in advance. There are still easy questions (not even involving coding, just reasoning about the problem and identifying edge cases) where LLMs fail spectacularly.

I am a Wire user. I am not happy that, if their server goes down, all the text and images that I shared with other users will become unavailable to me. As a special case, I am not able to look at my own old messages while using my laptop on the airplane, as the Linux client is just a webview. On the phone, it works.

Backups half-way solve the issue with text messages - I would still need to contact someone with sufficient development skills to decrypt the backup and extract the text in a readable form, but the information is there.

But with images, there is no recovery. And they apparently already lost some of my photos (the placeholder for some of them never gets replaced with the actual photo when I scroll up to year 2023).

Add to that the incompatible backup formats between the desktop and mobile apps.

So this is definitely not the claimed data sovereignty.

> So, those 5 golf players look better on the camera but their picture will be taken and shared with a phone 99.9999% of the times.

The point not in the article is that a phone camera, when used correctly (i.e., not defaulting to 1x zoom, maintaining a proper distance from the subject, and choosing wisely what's in the background), can produce photos of the same kind as a real camera in terms of composition. And that's what should be shared 99.9999% of the time. The extra resolution does not matter online, as most people will not zoom beyond what the platform, such as DeviantArt, offers as the default viewing experience - which is approximately 50% of the screen width, i.e., 1920 pixels on 4K displays.

Regarding the focal length difference, which is very well illustrated by the first two photographs, I would frame this point differently from what the article does.

Disclaimer: I don't have an iPhone 16, but my Poco X4 Pro 5G suffers from the same "fisheye" issue with its 25mm-equivalent lens. And I am not a professional photographer.

It is not a "bad camera" issue, but a composition issue forced by the short focal length second and the users' aversion to cropping the image first. You can easily avoid this issue by shooting at 2x digital zoom (or cropping) and going 2x further from the scene. And 2x zoom (with the corresponding decrease in the resolution of the resulting files) is how I shoot the majority of my photos.

So yes, I effectively have a 3MP camera, not 12MP (and not the hyper-marketed 108MP AI camera), and, for many purposes, it's still good enough.

P.S. Composition matters a lot. The right photo has the right proportions between the players and the trees in the background. The inclusion of the yellow golf flag in the "beginner photographer's" photo is also what makes the scene more complete artistically and worth hanging on the wall.

What's missing is a comparison with hiring someone to write that page of text. I.e., how much excess CO2 would a human produce writing the same page of text vs doing nothing for the same amount of time (and that's not zero - ask any freediver about their ways to reduce CO2 production in the brain), and so on.

The article title does not fully match the content. A more appropriate title would be: "In the issue tracker, write issues, not user stories".

The primary call to action here is to separate product-level details (including "user experience") from the issue tracker. The difference lies between the holistic view required for product-level work and the need for a clearly defined scope for tasks and issues.

Just guessing: the rules, as they are written, prohibit any usage of LLMs (not just for text generation), and, as is the case with any regulations, must be interpreted literally.

That's mine! And it's already fixed; I am indeed amazed by the developer's responsiveness.

8B (Gemma 3n)

Harper is so basic that I can't recommend it.

Does not catch a singular/plural discrepancy between the subject and the verb in a sentence--a common mistake when the expressed thought applies equally to one thing or to many things.

Does not catch a missed indefinite article--a common mistake for speakers of languages that don't have articles. Similarly, does not catch the use of the indefinite article for a thing already mentioned before.

Does not even catch the obvious "don;t" typo.

I don't know why 6500K is preferred here.

I think it doesn't make any difference. Back in Russia, I used 2700K or 3000K LEDs. Here in the Philippines, high-CRI warm-color LEDs are unobtainable, and the culture exhibits a nearly-universal shift to 6500K indoor lights (unlike in Europe), so I use 6500K, just like everyone else here. It still works.

Sorry for exposing my personal medical data, but... I literally cannot fall asleep without at least some light. So I sleep with lights on. Trying to be like normal people would only make my overall health worse.

Is this a medical condition that has a name?