Zak

The premium flashlight market has solved this with semistandard protected button-top cells.

These consist of a standard cell with a protection board stuck on the end, and usually a smaller raised positive terminal. This adds 3-5mm to the length, and temporarily breaks the circuit in the event of over-voltage, under-voltage, or over-current/short. There's no written standard, but most brands are making pretty much the same thing, and they almost always interchange. Safety does not require it to be proprietary.

Here's an example from a flashlight manufacturer: https://www.skilhunt.com/product/bl-135-3500mah-18650-protec...

Here's an example from a third-party battery distributor: https://illumn.com/18650-keeppower-p1840c-4000mah-protected-...

It's weird that rechargeable bike lights almost universally use soldered batteries. Conventional flashlights with onboard charging often use standardized cylindrical Li-ion cells that are field-replaceable by unscrewing the tailcap. They're usually 14500, 18350, 18650, or 21700 size, sometimes with a protection circuit stuck on the end (making it semi-standard, but easier to handle safely).

The light in the link uses an absolutely tiny 0.5Wh battery, which strikes me as insufficient for cycling. I use a light with a 21700 cell (20Wh).

Being a monopoly alone does not explain Google's behavior here. Google has a long history of not providing support.

The company's philosophy on the issue seems to be rooted in the hacker ethos of not giving repetitive work to humans if a machine could do it. When a machine can't do it, Google's thinking is to improve the machine. If the systems are good enough, nobody will need support.

The systems are not good enough, and that has been true for a long enough time that a reasonable company would have realized that a good support team is necessary until some far-future time.

Interestingly, conservative Republican sheriff Ross Teeple also spoke against the idea of networked/data-storing ALPRs and Flock in particular at that hearing.

Manual safety or not is a real debate among people who carry pistols for self defense and law enforcement, and no manual safety is the majority position. The reasoning is that an extra step when the user needs the pistol in an emergency is problematic, a properly designed pistol can only fire if the trigger is pulled, and a properly designed holster makes pulling the trigger impossible until the pistol is drawn. Nobody seriously advocates carrying an unloaded pistol for those purposes, and only lunatics advocate having people who don't know what they're doing carry pistols.

It's not a great analogy to Wordpress, which attempts to provide as much capability as possible in a CMS while requiring as little expertise from the user as possible. Vanilla Wordpress, kept up to date is pretty safe. Plugins are just a click away though, and using plugins safely requires evaluating each plugin's risk profile and track record individually, which is real sysadmin work.

Gnome's sharpest edge is that it has tried to simplify to the point that it doesn't have a UI for the thing I want to do. There might be an extension, and if I'm really lucky, it isn't broken and actually supports the current version.

I've been using it on my tablet, as I do find it to be a nice UI for the touchscreen, but I may switch to KDE because Gnome's onscreen keyboard is horrid, and difficult to replace.

I don't disagree in principle, though I question it in practice. I suspect network interfaces disappearing from KDE on mainstream distributions is rare. That may vary if you're using "Dave's Janky Arch Fork" or some such.

Commercial operating systems aren't immune. I found several people having a similar problem with Windows 11 on Microsoft Q&A, as well as a thousand third-party sites offering instructions to fix it, likely all infested with ads for "fix your computer" programs that will make things worse: https://learn.microsoft.com/en-us/answers/questions/5583596/...

I'd expect Mac OS to be better, in part because it only has to support a few models of computer. Perhaps it is, but it's also not immune: https://superuser.com/questions/1756912/ethernet-interface-d...

Everything is full of sharp edges. It's just a matter of which ones are more painful to you.

The author's inability to uninstall or fully disable AI features is one for Mac OS. Window management is a common one for people used to other systems.

Windows has many, with (again) AI features that are hard to get rid of, and attempts to coerce users to create an online account being two of the most prominent lately.

So yeah, KDE's NetworkManager frontend is a little janky sometimes, and they should fix that. Is it worse than trying to coerce the user into using services they don't want to? For some people, probably. For me, no.

I remember reading about narrow exceptions, but a quick search suggests I was mistaken.

Yes, there's a difference: US citizens have the right to enter the USA; visitors do not. US officials can deny a visitor entry arbitrarily, and the visitor has limited recourse to appeal such a decision. A visitor who does not cooperate with officials who want to search something is likely to be denied entry.

A citizen may only be denied entry if they're entering the country for the purpose of committing a crime, and if the government can prove that, they'll probably admit the citizen and immediately arrest them instead.

> It is insane to me that judges consistently rule that the constitution doesn't apply because there's a "border" within 100 miles.

People keep misconstruing this.

The border search exemption only applies when someone is actually crossing a border, and it does grant broad authority to search whatever they're carrying with them at the time. The authority to do border enforcement within a 100-mile zone around borders does not grant a broad authority to search. It does grant a narrow authority to do things like traffic checkpoints, but those don't permit searches for anything smaller than a person without additional evidence.

I do think the ruling establishing the authority to run checkpoints other than actual border crossings was wrongly decided and harmful, but it's not what's at issue in this case.

Most of them do. They're usually constructed as restrictions on the government's authority rather than grants of rights to a specific group of people, and there are court rulings clearly stating rights like freedom of speech and freedom from unreasonable search apply to any person subject to the government's authority.

Non-citizens don't get the right to bear arms as far as I know.

It might - let's assume it does. My point is that it's better to use the explicit optimized method for joining strings in a performance-sensitive context than to try to meet the conditions for an implicit optimization.

I find that decision a bit odd given that accumulating a string with a loop is also quadratic in Python if you use = instead of +=, or even if you use += when the left operand isn't provably unshared. I don't believe removing loops was seriously considered.

The footgun isn't `reduce` in particular, but failing to use `join`.

Nucleus programming language: https://github.com/zakwilson/nucleus/

It's still pretty early, possibly buggy, and almost certain to get breaking changes without warning.

My original thought was to combine Lisp syntax and structural macros with C semantics and see how much extra abstraction that would provide, but variants of that idea have been tried before. They may have been compelling at a certain point in time, but I think a new systems programming language in 2026 should offer more, so it does.

Some features include: strong typing, protocols (similar to Clojure or Rust traits), non-nullable references, error values, bound error handlers (a bit like Common Lisp), optional lexical lifetimes, namespaces, and no mandatory runtime overhead relative to C.

It's self-hosted, which I've noticed several new systems programming languages announced recently are not.

The brand name that comes to mind is the Wicked Lasers Flashtorch, which uses a 65W halogen bulb that emits most of its energy as infrared, and a ~30 Watt-hour Li-ion battery. The old-fashioned light source is better suited to party tricks than illumination.

More modern, and also capable of improvised cooking are handheld searchlights from Acebeam and Imalent with many large LEDs, cooling fans, and even bigger Li-ion batteries. There are third-party videos of these and the older halogen type cooking eggs and setting fire to paper on Youtube.

What you're describing from the ads obviously violates the laws of physics. An alkaline AA would have a hard time cooking an egg on a dead short.

> flashlights so powerful they'll fry an egg by shining at them

Flashlights that can do this exist. I don't know if they're being advertised on Meta platforms.

"May contain" should provide legal protection. It may be compliance tits in some cases, but that's better than forcing the addition of a compliance ingredient.

"May contain" labeling seems like the better solution to this problem. This should be allowed for allergens that are present in the manufacturing or packing facility but not intended to be ingredients in the product.

This seems like a beginner-level OOP-in-Python tutorial, and I'm disappointed the author didn't demonstrate a functional abstraction for `find_by`.

The term "missing person" calls to mind someone who is in serious danger, but a majority are something minor, like someone quitting a job without notice or taking an unannounced trip. The vast majority are resolved by finding the person alive.

Flock helping close 10K cases doesn't surprise me at all, but it doesn't meant those people wouldn't have been found without it. It may mean less work for the police, but the police don't usually put a lot of effort into a missing person without evidence they're in danger.

Google Maps has a better excuse for wanting to be an app than most since a primary use case on mobile is turn-by-turn navigation. That's possible to do in a web page now, but several features that improve the experience are not available in browsers.

The only time I've ever managed to burn through a quota that fast (on the cheap plan) was with an open-ended request to check a codebase for any defects or deficiencies. It dispatched five Fable subagents.

I also found the description of Lisp development weird, as if it's a second-hand repetition of someone else describing watching a Lisp developer in action.

The workflow I always use is to write code in my editor and use a keyboard shortcut to evaluate it. Typing directly into the REPL is for one-off state updates, inspection, experiments, and similar throwaway code.

This is a half-truth. In a full banana republic, technical compliance with the law will not prevent consequences for failing to do what the authorities want. In a jurisdiction with perfect rule of law, it always will. The USA is somewhere in between.

One of the laws that's enforced pretty well in the USA is the protection against unreasonable search. Most of the time, a search requires showing a judge evidence that the search is more likely than not to reveal evidence of a crime. Exceptions are narrow and specific; the government's options to punish someone who refuses to decrypt data at the border are limited to brief detention and seizure of the medium.

Not yet tested is the idea that erasing data on the spot satisfies the purpose of the border search exception, which is to prevent importation of things that are illegal to import. This case might address that question.

For exactly the border search scenario, I wish smartphones could be imaged and restored as easily as PCs. Imagine booting the phone from a flash drive, making an encrypted image of the phone on said drive, and writing a fresh OS before reaching the border.

There's no deception required to protect sensitive data or avoid the seizure of an expensive phone. Consent to unlocking the phone, refuse to unlock the drive. The drive gets seized and you go on your way (if you're a US citizen entering the USA).

Some time ago, Android with a custom recovery could come close to that, but it was fussy and as far as I know, no longer viable. Increased use of TPMs for storing credentials seems to be at least one of the reasons.

They do, and 20 years ago a move like that would have also drawn widespread condemnation.

It's not as big a problem as it is with phones though because the vast majority of PCs on the market are not locked down in this way, and only a handful of PC applications (a few competitive games) use remote attestation to check for an unmodified Microsoft OS.

Voiding a warranty due to unauthorized repair is only one of many things a company can do to make third-party repair, or any repair more difficult.

The first part is clearly illegal in the USA. Is it not in the EU?

https://en.wikipedia.org/wiki/Magnuson%E2%80%93Moss_Warranty...

Microsoft proposed locked-down PCs decades ago and got criticism from sources as mainstream as the New York Times. Everyone more technical than that, with the possible exception of corporate IT types flatly rejected it.

I'm sad there was so much less backlash to restricting user control on phones.

https://www.nytimes.com/2003/06/30/business/technology-a-saf...