I stopped reading at "The policy correlate of the vocabulary theft is equally clean." Please write your own blog posts; if I wanted to listen to Claude ponder about the crypto market I would have prompted it myself.

Writing tip: you do not need to have LLMs expand your ideas into a longer form for you. Spreading out your ideas across a longer post does not make them better.

No, but it satisfies a number of algorithms that drive traffic to websites better, and there are honestly people who will upvote something that starts strong and is long without reading the whole thing, who wouldn't upvote that same strong start if it wasn't padded with filler to make it look more exhausitive.

Not a fan either but there are real game theoretic reasons to do it.

You think this was written by AI? I'm curious, why? Anytime I'm about to read a long post I generally paste it into a few detectors for sanity, but none of them flagged anything for this one. I read the entirety and thought it was pretty compelling - long maybe, but thoughtful. I appreciated the historical quotes interjected, and didn't find the prose suspect.

"On the surface it looks like an engineering question, or a career planning question, but underneath it is an existential question:"

"This gives us the first corollary of end-state thinking cold but honest:"

"What does the end-state look like? My take is simple — a single line: PRD is Code."

"This shift brings a quiet, lethal corollary: "

"But think clearly about what letting go means."

"It sounds like a lament; it is actually a refining process. "

"When the tide of the times is pushing you along, slow can be a discipline."

These outtakes are all fairly obvious unedited claudeslop that add nothing to the argument. The excessive use of bolding and em-dashes throughout the article are also signals.

Some of the formatting for emphasis and em-dashes looked a bit iffy. It's hard to imagine AI wasn't involved, at least an an editor or to discuss ideas. To be clear, I don't think there's necessarily anything wrong with this.

"—" "blast radius"

done :)

I actually didn't smell AI at all while reading this article, because the idea density seemed high, rather than when AI just expands a few ideas into repetitive mumbo jumbo. The last portion was a bit wordy and metaphysical, but he landed the plane at the end.

Isn't the original post in Chinese? English is probably llm translated - like the whole blog

LLM translations of non-LLM text don't generally look as bad as this does. Many of the LLM-y traits -- inappropriate overuse of bolding and em-dashes, aggressively punchy sentences, and an essay that seems artifically stretched out to hit a word count -- are not affected by translation.

You can switch to the Chinese language version. Why don't you read that one first before deciding if it's LLM generated.

I enjoyed the fact that AI was used to deliver the anti-AI message.

that fewer people will read because it spends pages saying what could be paragraphs

Yeah, I got through about 2/3 of it, then called it quits.

I wonder if the author tried to sit down and read the whole post, from beginning to end in one go? The AI-written posts feel sludgy, like plain oatmeal that's been cooked too long, homogenous and bland.

Yeah it became very wearisome to read. I started skimming about halfway through.

Maybe you don't understand, if we didn't have LLMs expanding our ideas into longer form we wouldn't have anything to give to LLMs to summarize.

My boss pays me for the lines of code, right?

Glad I wasn't the only one!

"I would have written a shorter letter, but I did not have the time." - Blaise Pascal

Unfortunately with word generators this has become -

I would have written a letter, but I couldn’t be bothered.

third time i see this cliche quote in two days

The quoting will continue until writing improves

The thing is that writing - especially writing well - is fairly hard, and does take time and mental effort. So lazy authors (or wannabe authors) take the shortcut. But the fact remains that most of us do not want to read AI-generated slop, and human-written slop is not much better.

So there is a point to the quote, no matter how many times you've read it. But there is a countering point as well: If you don't have time, don't write. Don't use that as an excuse to get lazy and shove slop at us.

And yet people(?) keep publishing stuff that's too long.

> Fable might have been trained using Ellsworth's translation and as such it's very directly able to crib from it

The `cp` program on my computer also has the remarkable ability to produce a faithful translation of The Three Musketeers when provided one as input.

Not necessarily. If you are using macOS and APFS, it will just make a link, it won't actually make a copy.

Please write your own blog post. I couldn't read this because the LLM default tone is very grating.

Are there examples of the outputs the LLMs under test generated? I couldn't find any detailed ones in the paper or code.

The result here seems to be "Our Judge LLM gave another LLM a 21% grade for some code it generated", which is ... not qualitatively meaningful at all to me.

It's a temperature difference, not an absolute temperature. If it was 50F yesterday and 100F today, that means it was 10C yesterday and 38C today.

I wonder if it could have been done to shut up some static analysis tool?

That's my guess too. They were probably using ranqd1, then someone turned on a new warning, then someone else had to figure out what to do with the warning without really understanding the change they were making.

It'd be interesting to see if the same randomizer is used in other SMB games after NSMB and any ports.

I don't think this would make a warning go away. value + (value >> 32) is still a 64bit int value that gets then cast to a 32bit int.

Whoops, good point. Uh.. maybe there's a cast in the original source and the bit shift was added at the same time?

IDK, it doesn't seem ironic to me. Being the only store that sells something is pretty different from being the only store allowed in town.

They pay game companies to make sure they're the only store in town for popular new releases. There are other stores for other games, but not the Epic exclusives. Just like there are other stores for Android than Google Play, but lots of stuff uses Play Services and thus can't work without Play.

The obvious difference is that those game developers have the choice to be exclusive to the epic store, or steam, or neither or whatever. They are making the choice.

And game developers have a choice to not develop for iOS, but they do anyway.

It's not the same. Anyone who wants to develop a SaaS offering that requires a mobile app, which is basically everything nowadays, MUST take into account Apple's App Store and the Google Play Store because almost all of the target users will have those devices.

A game developer selling an exclusive to Epic is much different. They're not gatekeeping the platform which is the PC and no one is being forced to use the Epic store for distribution to get their games on the PC.

So to get access to 100% of the market on mobile platforms, you MUST deal with Apple AND Google. To get access to 100% of the market for anything on the PC platform you MAY deal with Epic, Steam, all of the above, or self-distribution. The difference is that developers get to choose which distribution strategy and platform makes the most sense for them in terms of cost, support, target audience, etc..

It's like if Walmart and Target were the only stores in North America and most buyers only shopped at one or the other. Would "choosing" not to sell your product at Walmart be a real option at that point?

The difference there is there is no other way to distribute iOS Apps. Period.

There are so many ways to distribute PC games and software, you cannot even begin to enumerate half of them.

Mostly the same on Android too. With a simple toggle, you can then install other App Stores like Amazon's, etc. Many phones come with alternative App Stores pre-installed too.

It depends how you break up the stats. The total number of officers feloniously killed is a little higher than the number of accidental deaths for recent years, but motor vehicle crashes is typically the largest single category. https://ucr.fbi.gov/leoka/2019/home

Yes, obviously? Of course they are going to get hurt doing their job. A lot of their job involves performing odd manuevers in traffic, for example. Do you have a suggestion for how to do policework without police officers ever having to accept the possibility of getting hurt?