nhubbard

It does let you run things as root.

I do recall that Fujitsu had a very... unique approach to exposing the Setup option in their firmware. It was just a UEFI boot entry. If that entry was removed, you couldn't get back into the firmware setup; you had to find the name of the setup EFI file and then run that from the EFI console.

What, people don't just recall those from memory? I fall asleep saying -Xmx and -Xms every night

Would love to see the source code for this and the underlying details like Classic or Carbon, and the libraries mentioned on Tinker Different for TLS, HTTP/2, and Unicode

Correct, but they stylized it as "eyePhone" (from MomCorp, the all powerful, caring conglomerate), and that episode is the origin of the famous "Shut up and take my money!" meme.

And something similar still happens if you install more than one drive in your machine with fresh Windows installations. Ugh.

I'd guess some constraint on their end related to the Zero Data Retention (ZDR) mode? Maybe the 1M context has to spill something onto disk and therefore isn't compliant with HIPAA.

That's what I figured. Some day eventually it will be possible. Until then, it's only LM Studio or Ollama as a potential hookup.

I've got some ideas inspired by this project. It's promising.

How practically could we drop in Apple Intelligence once it's using Gemini as its core for a 100% local AI agent in a box?

They did it in the 1990s and it failed so hard that it almost took down the company.

Maybe a better title would say "Asian nations [independently] roll out 4-day weeks, WFH to solve fuel crisis"?

This is the official Wikimedia Foundation status page for the whole of Wikipedia, so it's a reliable primary source.

I wouldn't be surprised either. But the original formatting of the worm makes me think it was human written, or maybe AI assisted, but not 100% AI. It has a lot of unusual stylistic choices that I don't believe an AI would intentionally output.

Wow. This worm is fascinating. It seems to do the following:

- Inject itself into the MediaWiki:Common.js page to persist globally, and into the User:Common.js page to do the same as a fallback

- Uses jQuery to hide UI elements that would reveal the infection

- Vandalizes 20 random articles with a 5000px wide image and another XSS script from basemetrika.ru

- If an admin is infected, it will use the Special:Nuke page to delete 3 random articles from the global namespace, AND use the Special:Random with action=delete to delete another 20 random articles

EDIT! The Special:Nuke is really weird. It gets a default list of articles to nuke from the search field, which could be any group of articles, and rubber-stamps nuking them. It does this three times in a row.

I will say that 26.4 beta 2 was the first time I've regretting using betas since Sonoma beta 2. The Sonoma beta ruined the firmware on my machine and Apple had to replace the logic board; the latest Tahoe beta broke all networking on my machine and I had to erase the installation to fix everything. I've since dropped off the beta train for the time being.

I already left the beta train on my iPhone because I had too many issues getting my grocery apps to allow me to place orders without going to my laptop and doing it in a web browser.

That's the one!

That's the right chip. The other comment shows off the article. I forgot that it was called the "sensor hub", that's why I couldn't find the post showing how it works.

I had never heard about this app. I thought the era of advertisements taking over the lock screen ended back in the Android 4.x days!

But also, thinking from the business perspective, it's difficult to make phones meet such a low price point without either significantly compromising their performance or stuffing them full of ads to subsidize the price.

While I generally agree with your sentiment, these tools aren't bad ones:

- Santa is a very common tool used by macOS admins to lock down binary and file access privileges for apps, usually on managed machines

- Disk Inventory X and GrandPerspective are well-known disk space usage tools for macOS (I personally use DaisyDisk but that requires a license)

- WizTree and WinDirStat are very common tools from Windows admin toolkits

The only one here I can say is potentially suspect is ClearDisk. I haven't used it before, but it does appear to be useful for specifically tracking down developer caches that eat up disk space.

One of the important reasons that it works so well is because it uses the Hexagon DSP in the Snapdragon processors to catch the events. That's why it's so hard to replicate. It's possible to do it entirely in software, but it chews through battery if you do it that way. I can't find it now, but there was an article a few years ago that explained how the feature worked.

And there's no way to program the DSP without being the creator of the device because Qualcomm requires DSP programs to be signed, as far as I'm aware, and the key has to be trusted by the device vendor.

> Hopefully the BIOS writers do not abuse this

Unfortunately, they have. Multiple examples from the excellent Cathode Ray Dude:

https://www.youtube.com/watch?v=q5M0TwnkWUM https://www.youtube.com/watch?v=ssob-7sGVWs

Dell likes to pull this stunt on other devices too. Like their 1L desktops in the OptiPlex line that I managed for many years. Even though we were using genuine Dell power adapters, if they became slightly unplugged but remained powered, they would enable PROCHOT.

This was fine until the machines randomly started setting PROCHOT on genuine power adapters that were fully plugged in. Eventually I just deployed a configuration with PDQ to all the machines that ran ThrottleStop in the background with a configuration that disabled PROCHOT on login.

Unfortunately, I couldn't get it to consistently disable PROCHOT pre-login, so students and teachers in my labs would consistently wait 3-4 minutes while the machines chugged along at 700 MHz as they prepared their accounts.

This wasn't PTC. It was repair lockouts instituted by the manufacturer of the trains based on a GPS geofencing beacon.

I've made a feature request there to add another GitHub Actions bot to auto-close issues reporting errors like this when an outage is happening. Would definitely help to cut through the noise.

https://github.com/anthropics/claude-code/issues/22848

I would normally do something like that; however, we opted to open an AppleCare Enterprise case for it, since it gets an engineer to look into the issue.

Both our Mac IT manager and I are thinking that it's the AV engine (Microsoft Defender) seeing fragments of YAML in the PSI tree and then blowing up after trying to parse the next line (which is not a truly valid YAML tree, it's a YAML tree with all the semantic details produced by IntelliJ). That in and of itself shouldn't be an issue, but I guess they weren't expecting random fragments of YAML to be copied and pasted.

We're still not sure why it only triggers in terminal emulators, though. That one is a total mystery.

I thought I was losing my mind for a second. What a strange script to run...

Update: Several reproduction sessions later, it appears that the content of one specific YAML file's PSI structure is causing a negative interaction with the system's EndpointSecurity extension process. Either one of the AV products on my work machine or the EndpointSecurity extension itself are panicking when the Pasteboard daemon attempts to output the contents. My work's IT team is awesome, so they upgraded a testing machine to 26.3b2, ran the same test, and it was also affected.

The issue was happening in Ghostty, iTerm2, and Terminal.app. It happened by (a) pasting directly into the terminal prompt with Command-V, (b) using `pbpaste` in any way (like piping it into `wc -c`), or (c) pasting into an editor like Neovim with Command-V.

It doesn't happen with any other file I tested it with. It also doesn't happen if I try to paste into another app, like Visual Studio Code, TextEdit, or IntelliJ IDEA itself.

Nothing is inherently wrong with the binary/text content of the file. It's all ASCII characters.

Since this is on Android, this policy should only apply to the version of the Messages app within the work profile, right? If it didn't and could access personal messages, that would be crossing a line.

Reading the post makes it sound like this only happens on managed devices; whether that means "owned and provided by work", "within the confines of the work profile on a BYOD devices", or both, I'm not 100% sure.

I had a very different reaction to this, because I just released a Rust library called ironbeam [0] without knowing anything about this. It was intended to be a pun on Apache Beam and the fact that iron can rust... I guess I'll have to live with it for now because it's way too high-effort to rename it after a stable release.

[0] https://github.com/nhubbard/ironbeam

Oh no, this is undoubtedly going to be terrible for the companies that built their OTT platforms on Vimeo. E.g. Dropout.tv.