cillian64

In case you didn’t know, there’s a lightspreader mod for the backlight which makes it a lot less terrible. I did it on my f91, the backlight still looks a bit odd but it’s much more readable

Thinking about it, even a lot of fancy fashion watch brands seem to do this, “Quartz”, “water resist 20m”, however many jewels. The old Casio watches are a bit worse for it though.

But check out the Casio MQ-24, it’s cheap and cheerful but has a much plainer face.

You could move the settings app to a less-used home-screen page, or you can remove the settings icon from the home screen entirely.

It's mildly annoying to have to do the swipe-down-and-search to get to settings but it's not an app I use all that often.

As someone who is usually on the other side of this interaction (software engineer at a company receiving bug reports), it's much easier when people write in their own words (even if they sometimes use the wrong terms or say something which can't be true) than when they use an LLM to make things clear and correct. I want to know what you're trying to do, what you expected to happen, and what actually happened - that doesn't benefit from having an LLM rewrite it.

Sure, I use precise and technical language, but I'm still a human and I can understand what people are getting at when they use imprecise terms. But if the text passes through an LLM I can't tell whether the result actually matches the original problem or whether the LLM has made guesses or filled in blanks.

A free claude account with no subscription gets you access to sonnet and I believe uses it by default over haiku

Generally the claude jargon is valid jargon I’ve seen real people use, it just uses jargon so much more often than a human would. And because it has particular jargon it likes, and claude is widely used, you see so much of that particular jargon that you get sick of it.

It’s like having one coworker with a very particular writing style which is mildly annoying, but then it suddenly feels like half the internet was written by that one person and it becomes a lot more annoying.

Yeh, I'm having some issues from the UK.

Looks like we're not just imagining it: https://fastmailstatus.com/cmt928hrt00gr0lqvevgyimt6

I think by "dangerous" they include things like "makes an edit to a config file outside the current project", not just "wipes the production database". So "dangerous" commands just means things we should ask the user for confirmation, not commands which definitely cause irreversible damage.

I changed from cursor to claude code a couple of weeks ago and this is the biggest difference I've noticed. cursor very rarely had to ask for permission for a command because it almost always used builtin find/read/edit tools. Whereas claude code makes extensive use of unix tools with mysterious flags which are basically impossible to whitelist and pretty impractical to manually review.

Overall claude code seems better at "just get things done" if you blindly trust it, but cursor was much better at keeping the human in the loop.

sed can be used to edit files (with the -i option). Not sure how this interacts with piping to its STDIN, but I wouldn't automatically assume `grep | sed` is safe.

I was thinking about this the other day, standard UNIX utilities seem poorly suited for this because they have so many weird options/flags with "unsafe" side effects. I have no idea if grep, for example, has some niche option to make it write/edit/delete files, so I wouldn't be comfortable blanket whitelisting it.

Even for extended workloads, a Pi 5 without a heatsink is still a hell of a lot faster than a Pi 4. And as sibling says, most users appreciate bursty speed while not doing prelonged compute (see also fanless laptops).

(Disclaimer, I work for raspberry pi ltd, not views of employer etc.)

You could say almost the same about Gorillaz except for the AI part.

I had that style of integrated panels installed on my (Victorian) house in the UK. The slates were old and crumbling so needed replacing and doing integrated panels meant only needing a fraction as much slate which saved a chunk of money. I think the result also looks a lot better than non-integrated panels, at least with slate roofs (which are common in my area)

Apparently integrated panels can be a little less efficient in hot weather as they don't get cooled by airflow under the panels and are less efficient when hotter. But it's a pretty minor effect, maybe a few percent of output. Seems like the best option on a new-build or if you're re-roofing anyway.

To look at it another way, just because some code I work on at my job is derived from open source MIT-licensed code doesn't mean I personally have the right to distribute it if my company doesn't want me to. I'd guess this comes under some generic "confidential information" clause in the employment contract.

It'll be the same as larger inverters for roof-top solar - they are constantly monitoring the mains cycle and will shut off if the voltage (or probably frequency too?) goes out of range, let alone drops out entirely. The relevant standards in the UK are G98/G99.

If you weren't aware, the HN guidelines now include:

> Don't post generated comments or AI-edited comments. HN is for conversation between humans.

Two advantages for me: It's nice that you don't break the connector if you trip over the cable or put the laptop down on a soft surface, and it's nice being able to charge while still using both USB-C ports (although I guess 3x USB-C would also solve that).

I don't really see any downside to a proprietary connector if you also have the option to charge over USB-C as well.

Isn't this exactly the same as flexoptix and FS have been doing for years?

The trick with neutral in the leaf is that you have to hold the gear selector in the neutral direction for about a second. No relation to brake pedal timing as far as I can tell. No idea why neutral has that delay given none of the other “gears” do.

One random preprint I found (https://arxiv.org/html/2505.09598v2) estimates inference is 90% of the total energy usage. From some googling around, everything seems to agree that inference dominates (or is at least comparable to) training on the large commonly-used models.

I was less surprised that inference dominates training after I read that chatgpt is serving billions of requests per day.

Matt Levine is “known for his humorous, witty, deadpan writing style” - I’m pretty sure that’s a joke.

> Because duolingo is designed for addiction

For people who have trouble keeping up hobbies, that's a feature. Even if duolingo isn't the ideal way to learn, it's a lot better than something I give up on or forget about after a week.

I have a load of these shell aliases as I spend a frankly ridiculous amount of time messing with git. `g` is `git status`, `d` is `git diff`, `gad` is `git add`, `ds` is `git diff --staged`, `gg` is `git grep`, `gbv` is `git branch -va`

Compared to the amount of code that people compile and the number of bugs seen and fixed in that code, that is a tiny number of bugs. I wouldn't say it's "never" a compiler error but when you find a bug in your program, it's almost certainly not the compiler's fault.

That article says that they only add copper in the handling/bridging parts of the panel, not to the actual finished PCB.

Other examples from the article but not in the headline:

Meanwhile, a separate AI summary told some users of the BBC Sport app that "Brazilian tennis player, Rafael Nadal, comes out as gay".

Yep, for me the reason to go with Hue is that the bulbs have good CRI, good colors, and the colors match between all the different types of fixtures. I've also not had a single Hue bulb fail in the 6 years I've been using them.

There is quite a bit of variation between SD cards, browser taking several seconds is surprisingly bad.

On Pi 4 I use a 2.5" SATA SSD (WD blue) on a startech USB3 adaptor and the performance seems to me much better than even good quality microsd cards.