toxik

There is no reason. I switched from iTerm 2 and while Ghostty is a lot faster and lighter, it has occasional rendering issues and does not detect URLs as well. It also doesn't have tmux control mode which I really liked.

I quit using iTerm 2 because of the AI features. Ew.

Cherry-picked is the term, not nitpicked.

I think autonomous mowers are more a tech problem. Just like the vacuum cleaners, they're only successful in the case of literally zero eventualities. Rake left out in the grass? Ope, guess you need a new mower blade. Dog poop or slick of jam on the floor? Congratulations, literal Picasso has been made for you. On the floor.

Ranting on, even dishwasher machines are like this. You have to stack and unstack, easily the most labor intensive part of the job, and you have to do it just right. Spatula jams the upper spray arm spindle? Sorry, best I can do is literally baked on residues that will NOT come off next cycle.

I'm not sure what point I was making but I'm sure this connects to shoddy LLM software in some way.

I think my point is that not everything requires a well-engineered solution. Many times a badly-made cheap solution is a better alternative than a well-made expensive solution. That is true of AI "art" too. If you need a picture for a throwaway slide deck, fine. If you're designing the cover of your new book, maybe not.

All of it is still important, but you don't _have_ to do it. The LLM will infill a generic solution that looks about as good as AI art. It will have bugs down the line as important levels of encapsulation were not observed, state was not thought about or formalized, and global data dependencies are implicited all over the place.

Yeah, this DIY MITM attack is entirely avoidable.

Is this sarcasm? You don't see a lack of something. That's what it means. To prove a reduction you have to look at aggregate statistics and sentiment analysis.

Agh I can't innovate when I have to [checks notes] put the ingredient amounts on the label!!!

This competition will be great for us consumers for a while. Until China undercuts everybody else for long enough to essentially monopolize the market. This pattern has repeated itself across many markets already.

And you just know Canonical thinks this is feasible because they can just copyright wash it with an LLM. "Write GNU coreutils in rust" is all you need to say.

Thanks kind stranger, genuinely appreciate the info. I had my suspicions.

How about: forecasting is something you can do by modeling, AI is just modeling with a computer.

Haha, oh the poor mega tech companies?! As if. Google sees a market, it wants to capture it. Same as Apple did.

I run iOS 18 and have no problems with my battery, it is at 73%. iPhone 14 Pro. Don't think I'm going to install 27 either.

$2000 is crazy talk

Yep, I wonder why the ANC socialt is so fragile. Turns me off from buying new ones.

What an absolute AI slop post.

This was very interesting, thank you for the link.

In program code? Unlikely, surely¡

I have noticed this too. I have been playing it quite a bit. There is no reason to have one unique solution per board though.

Some thoughts:

Insane is not harder than Hard. Often times the opposite seems true.

Scratch mode auto-committing on win is a bit too nice.

Dark mode is whack, terrible contrast.

I would like a player profile with stats and so on.

Random Insane seems to play like one of 50 levels. I recognize them often. Seems not so with hard. Now I just enter the puzzle IDs in turn...

Some puzzles seem to have no prior solutions as my time is often exactly the global average.

I would like to know how many solved a puzzle. Also if I did.

Overall great job and I love this. I would like it to be an app or something that can be bookmarked.

> Robot motion planning is basically just elevator scheduling with more degrees of freedom.

Certainly a take, ChatGPT. Not a very good one, but a take nonetheless.

Well it is different, you can have multiple solutions in this variant. Classical minesweeper is not like that. Imagine a grid with just a 1 and the rest is empty. There are then eight different solutions. The same situation can't happen in classical, it is simply going to reveal everything except the predetermined mine.

Because of this ambiguity, the search space is much larger here. Source : I play A LOT of Minesweeper.

... or cargo culted a site that did. I think this is far more common. It's almost like you're not a serious web site if you don't have a cookie popup of some kind.

I'm not sure why, but apparently my experience is unwelcome enough to be downvoted. Drive by changes are really annoying, IMO, and as the GP shows, the formatting tool outcome differs. So there are many "valid" versions of the same program logic. You have therefore not do what these formatters pretend to do, namely take this variable out of the equation. Rather, the wrong auto format can look much worse than the original.

I'm not sure why people felt formatters were necessary. but now they're part of the zeitgeist and there is no going back.

I have never discussed code formatting and drive-by changes more than in projects with auto formatters.

Well, they showed better performance if the AI had full state observation... Not very practical.

I think I would write mirrored. Especially in this day and age, there is no reason to write LTR over RTL by hand. You basically never have to write legibly for official reasons.

My big problem with the fountain pens is that the ink evaporates over a few weeks. Like bone dry, needs a refill.

Hard disagree on both accounts. Mechanical keyboards are obnoxious and tiring to use because of the insane travel. Fountain pens are nice when they work, but the ink evaporates, the nibs clog up, and accidentally bluing yourself is not uncommon among fountaineers.

Soon enough it'll read email [_]

_ https://en.wikipedia.org/wiki/Jamie_Zawinski#Zawinski%27s_La...