Rendered at 08:34:17 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
jstanley 22 hours ago [-]
This is really interesting but I wish LLMs wrote text that was a bit easier to understand.
> It requests the lethal puzzle exactly when all three conditions hold: 28 cases, including two ads instead of three, one ad closed early, a score 1 point short, and an ad worth just under 3.5×.
> Then we ran it live, in a test copy of the iPhone app running in Apple’s simulator, and read the rule’s inputs while we played. It had three good rewarded ads on record. The score was past the line. A rewarded ad from Google AdMob was loaded and ready. The rule priced it at 72.39, against 37.41 for the last one, which is 1.94 times. It needs 3.5. So on iPhone the rule is live, reading real ad prices, and waiting for an ad worth 3.5 times your last one.
What on earth is this supposed to mean?
truculent 19 hours ago [-]
Reading AI dialects feels similar (though somewhat easier) to reading Middle English text; the words individually make sense, but you’re much more aware of the communication medium and you have to concentrate a bit to pull out the meaning.
ErroneousBosh 17 hours ago [-]
I worked quite a bit for a company that had a lot of Asian developers in the early 2000s, and it reminds me a lot of textbook-Indian-English dialect.
The AI-dialect sounds very much like Harold, Clive, and Penelope from the Mumbai office; much of my work was then transcoding it into British English so that Surjit, Ismail, and Narinder from the Pollokshaws office could even begin to make sense of it.
(not the real names, but sufficiently similar)
thaeli 16 hours ago [-]
Huh. Maybe that’s why I find Claudish fairly easy to read.
joshspankit 18 hours ago [-]
My main thought while reading was “carefully crafting efficient, clean, and effective breakdowns is hard. I really appreciate that time and effort more now than I did before.”
jjbinx007 22 hours ago [-]
The page looked LLM made so I bothered to read I looked at the headings. Every one looked like nonsensical slop.
If they can't be bothered to write, I can't be bothered to read.
jgilias 21 hours ago [-]
You’re not supposed to read, you’re supposed to have your browser produce an AI summary of it lol.
cowlevel 17 hours ago [-]
That AI saummary will also be unreadable slop.
well_ackshually 20 hours ago [-]
You're doing really good. You're beating your high score. You've shown that you will watch ads because it gives you "rewards". Watching an ad after you failed ? You get to go back in the past and fix your mistakes. The next available ad is marked as particularly valuable by their ad network, paying them close to 4 times as much. They know you'd do anything to keep playing, even watching more ads.
It's the perfect time to put out a puzzle that you will almost instantly fail.
jstanley 20 hours ago [-]
I understood that much from the rest of the article, but I don't understand specifically the part I quoted. There is a lot of detail in there that doesn't seem to have made it into your translation.
jwr 23 hours ago [-]
I think it is so sad that humanity puts so much effort into things like this.
I thought by the 2020s we would have fusion power, and be busy exploring and colonizing space.
Instead, we got greedy corporations with moral-free executives (remember the "don't be evil" declarations? good times, huh) and incredible amounts of developer effort being poured into useless garbage.
chasd00 18 hours ago [-]
We’re coming up on the _second_ generation of the best and brightest programmer minds being wasted on getting ad clicks.
havaloc 18 hours ago [-]
If you're playing an ad supported game, you kind of know the deal already. It still is sad of course, but it is what it is.
Jaxkr 1 days ago [-]
The fact that impossible puzzles are generated to bait rewarded-ad views is something I've always suspected in mobile games, but I dismissed it as a skill issue on my end.
db48x 16 hours ago [-]
Why would you do that? It’s quite well documented in games that use microtransactions or advertising.
collingreen 4 hours ago [-]
Folks aren't generally expecting this kind of thing in my personal experience either. I wonder if it's something about people not expecting lies/deceit/abuse they haven't personally thought of before. We all can consider lying or stealing or even picking a pocket but very few consider massive, sociopathic abuse and corruption so it doesn't trigger that suspicion sometimes.
I cannot get people in my life to take any digital privacy seriously. A good example is when I couldn't convince a friend about the LG TV sampling your screen and phoning home, even with all the articles recently. I think this is why we like magic, fall for scams, and don't demand more from our politicians.
quikoa 1 days ago [-]
> Apple does let apps download JavaScript updates like this, which is how tools like Expo Updates work, as long as the update doesn’t change what the app is. Whether changing your pieces based on ad prices fits that is up to Apple.
But there is no way for Apple to verify "as long as the update doesn’t change what the app is" right?
user43928 22 hours ago [-]
I heard that with the latest policy update last week, Apple rolled out some heuristics that aim to detect feature flags that bypass App Review.
Supposedly this can also fire for legitimate A/B tests or other debug flags that aren't stripped from the Release build.
I suspect these games will be fine as long as they have an Account manager and pay Apple millions in commissions or for paid acquisition.
asafel 23 hours ago [-]
[flagged]
MDTHLN 1 days ago [-]
> The lethal puzzle is a generator, and any rule can ask for it.
Ashamed of myself for getting halfway through the article before realising it was AI written.
iamacyborg 1 days ago [-]
Hah, you can literally tell from the subheader.
> Rules downloaded from their server decide your pieces, your score and your ads. One waits for an expensive ad, then asks for a hand you can’t place.
wafflemaker 1 days ago [-]
I wouldn't be able to tell. No load bearing or is not x, it's y.
I actually don't mind if it's good enough, informative and not overly long.
Haven't really read anything super well written (in terms of language and form) lately, other than a really fun hn comment summarizing Trump's War.
But I just haven't read that much lately.
swiftcoder 24 hours ago [-]
The tell here is the "One ..., then ... <negative>". This is "not X, it's Y" with the clauses reordered.
For some reason the current crop of LLMs are really insistent on generating contrast between two things in every statement.
yen223 23 hours ago [-]
The problem is that the "it's not X it's Y" construction is very useful, and shows up a lot in human-written text.
swiftcoder 15 hours ago [-]
I think its really not all that common outside of Accenture pitch decks.
drstewart 22 hours ago [-]
>The tell here
Is this AI written? It has "this is the smoking gun" vibes, just rewritten.
teraflop 23 hours ago [-]
I don't know precisely what to call it, but there is something about AI's word choice and phrasing that really stands out to me.
It's like it chooses "simple" words, but uses them in unclear or fuzzy or slightly abnormal ways that make the overall meaning more difficult to parse. On its own, each instance is defensible, but the aggregate result when it happens many times over is very tiring to read. It reminds me of XKCD's "Up Goer Five": https://xkcd.com/1133/
Like this:
> One waits for an expensive ad, then asks for a hand you can’t place.
It reads as though it started with a straightforward idea, and then was forced to express it with words of only one syllable.
anon7000 13 hours ago [-]
This line stands out to me:
> One waits for an expensive ad, then asks for a hand you can’t place
Something about “asks for a hand you can’t place” just triggers the alarm bell. I think it’s unfortunately because I’m reading a lot of AI at $job. I do read books in my free time but something about the patterns…
Some more quotes from early on:
“Most of Block Blast isn’t in the app you install”
And then
“It waits for three things to line up: you finished your last three rewarded ads, the ad loaded right now is worth at least 3.5 times your last one, and you’re 1,000 points past your high score. All three get checked in one function.”
A really big AI tell for me is sensationalism and drama about things that don’t matter. Ok, fair, checking code in one function doesn’t seem dramatic. But consider this: a paragraph is building to a conclusion. You have a fairly long sentence, then a short sentence. The contrast in length gives the short sentence a lot of emphasis when I read it. In a newscaster voice: “all three get checked.. in one function”
But obviously it doesn’t matter at all, even to the technical reader, if they check this in one function or multiple functions. AI does this ALL the time, they emphasize some implementation detail that’s true, but completely irrelevant to the point the author wants to make. In fact this detail really should be removed from the article completely because it adds LITERALLY nothing to our understanding of how this app works.
So my alarm bells go off because a technical writer probably wouldn’t emphasize that point.
I’m not actively analyzing that, but when I read AI, I notice emphasis on things that are entirely irrelevant. The text starts feeling self-important about complete nonsense.
I think it all adds up for me, and I just notice the patterns in a surprising number of blogs and YT videos. It gets distracting, but IMO the bigger issue is that the wrong details are put in focus, so explanations get hard to follow.
iamacyborg 12 hours ago [-]
Claude loves announcing counts as well.
“X things follow”, or from your example, “It waits for Y things to line up”.
iamacyborg 18 hours ago [-]
I've read too much Claude slop. It's very easy to spot the sentence construction.
cowlevel 17 hours ago [-]
The more AI slop you read the more you'll notice its patterns.
abound 19 hours ago [-]
Same. The underlying story is interesting and horrifying (but also unsurprising), but the LLM write-up sucks all the air out of it.
I want to hear about the human author's motivations for doing this, the hard parts of doing the deobfuscation, their emotions when finding out what it did, etc. This is none of that.
Freak_NL 23 hours ago [-]
I started reading on the assumption it was posted here because it was interesting, but this paragraph cinched it:
> 1. You finished your last three rewarded ads. Every time a rewarded ad starts, the rule pushes false onto a list and keeps the latest three (lines 174–183). If the ad SDK reports that the reward was earned, it flips that entry to true (191–199). The check needs all three to be exactly true:
Pointlessly spun out prose. Pointless headers everywhere. Pointless lists. Yes, we understand what a list of booleans is, and that the game looks for a state where the last three are true. No need to spell it out like that.
Why do people post crap like this? Why do people post it to Hacker News? There is some merit to dissecting an attention sink ad-driven game, but this article just makes me feel like my attempt to grasp the meaning and writer's views fails every time — like I'm trying to grasp stalks of grain in a field, but my hand keeps passing through them as if they were holograms.
I would very much like to read people's thoughts with a forbearance for awkward prose, bad spelling, or suboptimal outlining, instead of that I get rewarded by text where the human who prompted it is suddenly absent.
Fuck this new AI world.
wodenokoto 22 hours ago [-]
While lots of parts of the article isn’t written well, the subject and what is uncovered is very interesting.
That is why it is posted.
GrinningFool 20 hours ago [-]
I think the bar should be higher. Incomprehensible is not justified by an interesting topic.
I wonder how such games can bring in millions per month.
Perhaps when your ads outperform the competition due to these devious optimizations, you can afford a higher bid for paid acquisition, keeping users coming.
noisy_boy 18 hours ago [-]
I play block blast classic mode with adguard and don't see any ads.
mzajc 22 hours ago [-]
I only forced myself to read through this slop prose because I found the topic interesting, but I'd much rather be reading something that doesn't repeat itself constantly and sounds human.
peter_d_sherman 7 hours ago [-]
>"The lethal puzzle is a generator, and any rule can ask for it. The ad rule is one caller. While we were testing it live on iPhone, a second caller fired on camera: PersonalMarginBreakTrait. Once you’re 10% past your high score, it rolls on every deal, and 20% of the time it asks for the same lethal puzzle:
Our record was 1,304, so its line was 1,434. At 3,356 points it rolled, and the game dealt a five-long bar, a five-tall bar and a zigzag. The game’s own record of that deal names this rule as the caller, shows algorithm 9 both requested and used, and labels all three pieces 致死题, lethal puzzle."
I love it!
From a linguistic "this one takes the cake in terms of strange terminology" strange terminology, this one takes the cake!
I could see this becoming a meme on the Internet... just like "pwned" (misspelling of "owned") was a long time ago...
I almost want to write a sitcom episode around it... something like "you've been playing this game too long... so now you get a "lethal puzzle!"..." :-). Classic! :-)
Anyway, very interesting article!
MisterMunchkin 23 hours ago [-]
Holy shit that is evil.
void-sky 23 hours ago [-]
Why is vibration a parameter?
Jare 18 hours ago [-]
If metrics have found any statistical difference among people with vibration activated or not, then there's a value in knowing. The game refuses to vibrate on my phone (no idea why; Duolingo vibrates fine) so I can't say what they do with it or not, perhaps they have "soothing" type vibrations my phone doesn't support, that have an impact on your emotional reactions (statistically speaking).
nova-47 22 hours ago [-]
[flagged]
wodenokoto 1 days ago [-]
I’m a bit unsure why they want to change gameplay based on the value of the ad.
It’s not like I get sucked into the game and decides to watch more expensive ads (which is the mechanic of golden berries or whatever)
As far as I understand the rule, high value players are more likely to die - but why?
Timshel 24 hours ago [-]
They want to ensure you watch it.
If you tend to watch ads and have a high score then when killed it's likely you will watch the ad to revive.
wodenokoto 24 hours ago [-]
I guess there’s just a lot I don’t know about this ad market. Is the high value ad only available to show for a limited time?
db48x 21 hours ago [-]
Very likely yes. Consider an ad for a political candidate: before a particular date it is worth a lot, after that date it is worthless.
ImPostingOnHN 15 hours ago [-]
> It’s not like I get sucked into the game and decides to watch more expensive ads
It seems to be exactly like that for a sufficient portion of their users.
wodenokoto 12 hours ago [-]
Explain how the user makes the decision. It's not like they click "watch expensive add and gain 2 coins, or watch normal add for 1 coin" The user doesn't know what ad they are about to be shown and has no choice in getting and ad that can revive their play.
ImPostingOnHN 8 hours ago [-]
when a user has lost, but can watch an ad to "revive" and continue the game, they are more likely to sit through that ad placement
some advertisers are thus more likely to pay more for those placements, and thus the placement is paired to one of these higher bidding campaigns
> It requests the lethal puzzle exactly when all three conditions hold: 28 cases, including two ads instead of three, one ad closed early, a score 1 point short, and an ad worth just under 3.5×.
> Then we ran it live, in a test copy of the iPhone app running in Apple’s simulator, and read the rule’s inputs while we played. It had three good rewarded ads on record. The score was past the line. A rewarded ad from Google AdMob was loaded and ready. The rule priced it at 72.39, against 37.41 for the last one, which is 1.94 times. It needs 3.5. So on iPhone the rule is live, reading real ad prices, and waiting for an ad worth 3.5 times your last one.
What on earth is this supposed to mean?
The AI-dialect sounds very much like Harold, Clive, and Penelope from the Mumbai office; much of my work was then transcoding it into British English so that Surjit, Ismail, and Narinder from the Pollokshaws office could even begin to make sense of it.
(not the real names, but sufficiently similar)
If they can't be bothered to write, I can't be bothered to read.
It's the perfect time to put out a puzzle that you will almost instantly fail.
I thought by the 2020s we would have fusion power, and be busy exploring and colonizing space.
Instead, we got greedy corporations with moral-free executives (remember the "don't be evil" declarations? good times, huh) and incredible amounts of developer effort being poured into useless garbage.
I cannot get people in my life to take any digital privacy seriously. A good example is when I couldn't convince a friend about the LG TV sampling your screen and phoning home, even with all the articles recently. I think this is why we like magic, fall for scams, and don't demand more from our politicians.
But there is no way for Apple to verify "as long as the update doesn’t change what the app is" right?
Supposedly this can also fire for legitimate A/B tests or other debug flags that aren't stripped from the Release build.
I suspect these games will be fine as long as they have an Account manager and pay Apple millions in commissions or for paid acquisition.
Ashamed of myself for getting halfway through the article before realising it was AI written.
> Rules downloaded from their server decide your pieces, your score and your ads. One waits for an expensive ad, then asks for a hand you can’t place.
I actually don't mind if it's good enough, informative and not overly long.
Haven't really read anything super well written (in terms of language and form) lately, other than a really fun hn comment summarizing Trump's War.
But I just haven't read that much lately.
For some reason the current crop of LLMs are really insistent on generating contrast between two things in every statement.
Is this AI written? It has "this is the smoking gun" vibes, just rewritten.
It's like it chooses "simple" words, but uses them in unclear or fuzzy or slightly abnormal ways that make the overall meaning more difficult to parse. On its own, each instance is defensible, but the aggregate result when it happens many times over is very tiring to read. It reminds me of XKCD's "Up Goer Five": https://xkcd.com/1133/
Like this:
> One waits for an expensive ad, then asks for a hand you can’t place.
It reads as though it started with a straightforward idea, and then was forced to express it with words of only one syllable.
> One waits for an expensive ad, then asks for a hand you can’t place
Something about “asks for a hand you can’t place” just triggers the alarm bell. I think it’s unfortunately because I’m reading a lot of AI at $job. I do read books in my free time but something about the patterns…
Some more quotes from early on:
“Most of Block Blast isn’t in the app you install”
And then
“It waits for three things to line up: you finished your last three rewarded ads, the ad loaded right now is worth at least 3.5 times your last one, and you’re 1,000 points past your high score. All three get checked in one function.”
A really big AI tell for me is sensationalism and drama about things that don’t matter. Ok, fair, checking code in one function doesn’t seem dramatic. But consider this: a paragraph is building to a conclusion. You have a fairly long sentence, then a short sentence. The contrast in length gives the short sentence a lot of emphasis when I read it. In a newscaster voice: “all three get checked.. in one function”
But obviously it doesn’t matter at all, even to the technical reader, if they check this in one function or multiple functions. AI does this ALL the time, they emphasize some implementation detail that’s true, but completely irrelevant to the point the author wants to make. In fact this detail really should be removed from the article completely because it adds LITERALLY nothing to our understanding of how this app works.
So my alarm bells go off because a technical writer probably wouldn’t emphasize that point.
I’m not actively analyzing that, but when I read AI, I notice emphasis on things that are entirely irrelevant. The text starts feeling self-important about complete nonsense.
I think it all adds up for me, and I just notice the patterns in a surprising number of blogs and YT videos. It gets distracting, but IMO the bigger issue is that the wrong details are put in focus, so explanations get hard to follow.
“X things follow”, or from your example, “It waits for Y things to line up”.
I want to hear about the human author's motivations for doing this, the hard parts of doing the deobfuscation, their emotions when finding out what it did, etc. This is none of that.
> 1. You finished your last three rewarded ads. Every time a rewarded ad starts, the rule pushes false onto a list and keeps the latest three (lines 174–183). If the ad SDK reports that the reward was earned, it flips that entry to true (191–199). The check needs all three to be exactly true:
Pointlessly spun out prose. Pointless headers everywhere. Pointless lists. Yes, we understand what a list of booleans is, and that the game looks for a state where the last three are true. No need to spell it out like that.
Why do people post crap like this? Why do people post it to Hacker News? There is some merit to dissecting an attention sink ad-driven game, but this article just makes me feel like my attempt to grasp the meaning and writer's views fails every time — like I'm trying to grasp stalks of grain in a field, but my hand keeps passing through them as if they were holograms.
I would very much like to read people's thoughts with a forbearance for awkward prose, bad spelling, or suboptimal outlining, instead of that I get rewarded by text where the human who prompted it is suddenly absent.
Fuck this new AI world.
That is why it is posted.
I wonder how such games can bring in millions per month.
Perhaps when your ads outperform the competition due to these devious optimizations, you can afford a higher bid for paid acquisition, keeping users coming.
Our record was 1,304, so its line was 1,434. At 3,356 points it rolled, and the game dealt a five-long bar, a five-tall bar and a zigzag. The game’s own record of that deal names this rule as the caller, shows algorithm 9 both requested and used, and labels all three pieces 致死题, lethal puzzle."
I love it!
From a linguistic "this one takes the cake in terms of strange terminology" strange terminology, this one takes the cake!
I could see this becoming a meme on the Internet... just like "pwned" (misspelling of "owned") was a long time ago...
I almost want to write a sitcom episode around it... something like "you've been playing this game too long... so now you get a "lethal puzzle!"..." :-). Classic! :-)
Anyway, very interesting article!
It’s not like I get sucked into the game and decides to watch more expensive ads (which is the mechanic of golden berries or whatever)
As far as I understand the rule, high value players are more likely to die - but why?
If you tend to watch ads and have a high score then when killed it's likely you will watch the ad to revive.
It seems to be exactly like that for a sufficient portion of their users.
some advertisers are thus more likely to pay more for those placements, and thus the placement is paired to one of these higher bidding campaigns