Knew it was getting bad, but Meta's facebookexternalhit bot changed their behavior recently.
In addition to pulling responses with huge amplification (40x, at least, for posting a single Facebook post to an empty audience), it's sending us traffic with fbclids in the mix. No idea why.
They're also sending tons of masked traffic from their ASN (and EC2), with a fully deceptive UserAgent.
The weirdest part though is that it's scraping mobile-app APIs associated with the site in high volume. We see a ton of other AI-training focused crawlers do this, but was surprised to see the sudden change in behavior on facebookexternalhit ... happened in the last week or so.
Everyone is nuts these days. Got DoSed by Amazonbot this month too. They refuse to tell me what happened, citing the competitive environment.
After 2 minutes at 150 kHashes on mobile, I finally see the first pixel of the progress bar filling up. Seems like it will take hours or a day to finish. Some estimate would have been nice.
Ironically I used a LLM to write a bypass for this ridiculous tool, doing hashing in a browser makes no sense, Claude's very bad implementation of it in C does tens of megahash a second and passes all of the challenges nearly instantly. It took about 5 minutes for Claude to write that, and it's not even a particularly fast implementation, but it beats the pants off doing string comparisons for every loop in JavaScript which is what the Anubis tool does.
for (; ;) {
const hashBuffer = await calculateSHA256(data + nonce);
const hashArray = new Uint8Array(hashBuffer);
let isValid = true;
for (let i = 0; i < requiredZeroBytes; i++) {
if (hashArray[i] !== 0) {
isValid = false;
break;
}
}
It's less proof of work and just annoying to users, and feel good to whoever added it to their site, I can't wait for it to go away. As a bonus, it's based on a misunderstanding of hashcash, because it is only testing zero bytes comparison with a floating point target (as in Bitcoin for example), the difficulty isn't granular enough to make sense, only a couple of the lower ones are reasonably solvable in JavaScript and the gaps between "wait for 90 minutes" and "instantly solved" are 2 values apart.
>It's less proof of work and just annoying to users, and feel good to whoever added it to their site,
this is being disproved in the article posted:
>And so Anubis was enabled in the tar pit at difficulty 1 (lowest setting) when requests were pouring in 24/7. Before it was enabled, it was getting several hundred-thousand requests each day. As soon as Anubis became active in there, it decreased to about 11 requests after 24 hours, most just from curious humans.
apparently it does more than annoying users and making the site owner feel good (well, i suppose effective bot blocking would make the site owner quite good)
It doesn't matter if your hottest loop is using string comparisons, as another poster pointed out in C you aren't even doing the majority of the second hash because you know the result (or enough of it) before finishing it. The JavaScript version just does whole hashes and turns them into a Uint8Array, then iterates through it.
Maybe post your brilliant solution to commercial companies with hundreds of millions in funding unrestrained bot scraping the Internet for AI training instead of complaining about people desperate to rein it in as individuals.
Anybody can prompt Claude to implement this, which was my point, it doesn't stop bots because a bot can literally write the bypass! My prompt was the proof of work function from the repository, asked it to make an implementation in C that could solve it faster, and that was about it.
This is fallacious and extremely disrespectful (or even malicious?). You don't have to propose a way to fix a broken thing to point out that it's broken.
Normal and sane people understand this intuitively. If someone goes to a mechanic because their car is broken and the mechanic says "well, if you can tell that you car is broken, then you should be able to figure out how to fix it" - that mechanic would be universally hated and go out of business in months. Same thing for a customer complaining about a dish made for them in a restaurant, or a user pointing out a bug in a piece of software.
> Anubis uses a Proof-of-Work scheme in the vein of Hashcash
And if you look up Hashcash on Wikipedia you get https://en.wikipedia.org/wiki/Hashcash which explains how Hashcash works in a fairly straightforward manner (unlike most math pages).
I'm surprised at the effectiveness of simple PoW to stop practically all activity.
I'll implement Anubis at low difficulty for all my projects and leave a decent llms.txt referenced in my sitemap and robots.txt so LLMs can still get relevant data for my site while.keeping bad bots out. I'm getting thousands of requests from China that have really increased costs, glad it seems the fix is rather easy.
The default is to allow non-Mozilla user agents so that existing (good) automation continues to work and so that people stopped threatening to burn my house down. Lovely people in the privacy community.
> Before it was enabled, it was getting several hundred-thousand requests each day. As soon as Anubis became active in there, it decreased to about 11 requests after 24 hours
I love experimental data like this. So much better than gut reaction that was spammed when anubis was just introduced
> These bots are almost certainly scraping data for AI training; normal bad actors don't have funding for millions of unique IPs thrown at a page. They probably belong to several different companies. Perhaps they sell their scraped data to AI companies, or they are AI companies themselves. We can't tell, but we can guess since there aren't all that many large AI corporations out there.
Is the theory here that OpenAI, Anthropic, Gemini, xAI, Qwen, Z.ai etc are all either running bad scrapers via domestic proxies in Indonesia, or are buying data from companies that run those scrapers?
I want to know for sure. Who is paying for this activity? What does the marketplace for scraped data look like?
So the elephant in the room: How much of HN is bot generated? Those who know have every incentive not to share and those who don't have no way to figure it out. At this point i have to assume that every new account is a bot
The commenters here don't care what the article is about when they can't access the article and the much more concerning question not about web scraping is.
I felt a vibe change, some are obvious and some not, but it does feel different, the main change i've seen is in downvotes, I don't say very controversial things and have had many things very quickly downvoted, and then slowly upvoted, I think hn was very slow to downvote in the past (except obvious trolls/spam). So for me the main worry is not even the comments, but the invisible bias generated by voting.
What is the point of these anti bot measures if organic HN traffic can nuke your site regardless? If this is about protecting information from being acquired by undesirable parties, then this site is currently operating in the most ideal way possible.
The information will eventually be ripped out. You cannot defeat an army with direct access to TSMC's wafer start budget and Microsoft's cloud infrastructure. I would find a different hill to die on. This is exactly like the cookie banners. No one is winning anything here. Publishing information to the public internet is a binary decision. If you need to control access, you do what Netflix and countless others have done. You can't have it both ways.
An interesting and sad aspect of the war on bots and scraping that is being waged is that we are hurting ourselves in the process, too. Many tasks I'm trying to get my AI assistant to do cannot be done quickly, because sites defensively prohibit access to their content. I'm not scraping: it's my agent trying to fetch a page or two to perform a task for me (such as check pricing or availability).
You aren’t scraping for the sake of training a model, but scraping the prices and availability is still scraping, right?
I think some of the folks running sites would rather have you go to the site and view the items “suggested based on your shopping history” (I consider these ads, the vendors might disagree), etc.
I’m more sympathetic to the people running sites than the LLM training scrapers, but these are two parties in a many-party game and neither one is perfectly aligned with users.
I'm getting this patern a lot on Prestashop websites, where thousand, to not say hundreds of thousand, of request are coming from bots not announcing themselves in the User-agent, and coming from different IP's
Very annoying. And you can't filter them because they look like legitimate trafic.
On a page with differents options (such as color, size, etc...) they'll try all the combinaisons, eating all the ressources.
I think that must be the point they're trying to make, yes
It also drives home that Anubis needs a time estimate for sites that don't use Anubis as a "can you run javascript" wall but as an actual proof of work mechanism that it purports to be its main mechanism
It shows a difficulty of "8" with "794 kilohashes per second", but what does that mean? I understand the 8 must be exponential (not literally that 8 hashes are expected to find 1 solution on average), but even as a power of 2, 2^8=256 I happen to know by heart, so thousands of hashes per second would then find an answer in a fraction of a second. Or if it's 8 bytes instead of bits, then you expect to find a solution after like 8 million hashes, which at ~800k is about ten seconds. There is no way to figure out how long the expected wait is even if you understand all the text on the page (which most people wouldn't) and know some shortcuts to do the mental math (how many people know small powers of 2 by heart)
We started building out a set of spam/fraud/bot management tooling. If you have any decent infrastructure in place already, this is a pretty manageable task with a mismash of techniques. ASN based blocking (ip lookup databases can be self hosted and contain ASN) for the obvious ones like alibaba etc, subnet blocking for the less obvious (see pattern, block subnet, alleviates but doesn't solve problems).
If you have a logging stack, you can easily find crawler/bot patterns, then flag candidate IP subnets for blocking.
It's definitely whackamole though. We are experimenting with blocking based on risk databases, which run between $2k and $10k a year depending on provider. These map IP ranges to booleans like is_vpn, is_tor, etc, and also contain ASN information. Slightly suspicious crawling behavior or keyword flagging combined with a hit in that DB, and you have a high confidence block.
All this stuff is now easy to homeroll with claude. Before it would have been a major PITA.
Worse than I could imagine? I imagine that bots might destroy the internet. Not just the internet as we know it; I mean make the internet completely unusable to any human being.
My grad school research was on computational models of human/machine cognition, and I'm now commercializing it as a 'proof-of-human API' for bot detection, spam reduction, and identity verification.
One of the mistakes people assume is that AI capability means humanness. If you know exactly where to look, you can start to identify differences between improving frontier models and human cognition.
One concrete example from a forthcoming blog post of mine:
[begin]
In fact, CAPTCHAs can still be effective if you know where to look.
We ran 75 trials -- 388 total attempts -- benchmarking three frontier AI agents against reCAPTCHA v2 image challenges. We looked across two categories: static, where each image grid is an individual target, and cross-tile challenges, where an object spans multiple tiles.
On static challenges, the agents performed respectably. Claude Sonnet 4.5 solved 47%. Gemini 2.5 Pro: 56%. GPT-5: 23%.
On cross-tile challenges: Claude scored 0%. Gemini: 2%. GPT-5: 1%.
In contrast, humans find cross-tile challenges easier than static ones. If you spot one tile that matches the target, your visual system follows the object into adjacent tiles automatically.
Agents find them nearly impossible. They evaluate each tile independently, produce perfectly rectangular selections, and fail on partial occlusion and boundary-spanning objects. They process the grid as nine separate classification problems. Humans process it as one scene.
The challenges hardest for humans -- ambiguous static grids where the target is small or unclear -- are easiest for agents. The challenges easiest for humans -- follow the object across tiles -- are hardest for agents. The difficulty curves are inverted. Not because agents are dumb, but because the two systems solve the problem with fundamentally different architectures.
Faking an output means producing the right answer. Faking a process means reverse-engineering the computational dynamics of a biological brain and reproducing them in real time. The first problem can be reduced to a machine learning classifier. The second is an unsolved scientific problem.
The standard objection is that any test can be defeated with sufficient incentive. But fraudsters weren't the ones who built the visual neural networks that defeated text CAPTCHAs -- researchers were. And they aren't solving quantum computing to undermine cryptography. The cost of spoofing an iris scan is an engineering problem. The cost of reproducing human cognition is a scientific one. These are not the same category of difficulty.
>The first problem can be reduced to a machine learning classifier. The second is an unsolved scientific problem.
I can't believe people are still using this as a generic anti-AI argument even though a decade ago people were insisting that there's no way AI can have the capabilities that frontier LLMs have today. Moreover it's unclear whether the gap even exists. Even if we take the claim that the grid pattern is some sort of fundamental constraint that AI models can't surpass, it doesn't seem too hard to work around by infilling the grids pattern and presenting the 9 images to LLMs as one image.
Everybody says that bots put websites down, while marketing oriented folks start practicing AO (agent optimization) - to make their offerings even more available and penetrating.
Taking a 2024 report on bot loads on the Internet is like taking a 1950s Car & Driver article for modern vehicle stats.
That’s how fast the landscape is changing.
And remember: while the report might have been released in 2024, it takes time to conduct research and publish. A good chunk of its data was likely from 2023 and earlier.
This recently registered account has been spamming links to their app in HN comments. They tried to post a Show HN as their first post after registering. Now they’ve tried to spam links to it in two comments (one dead) in this comment section.
I don’t think the comments are bot-written, but it’s ironic that when I first clicked on this comment section 2 out of 3 comments (including the showdead one) were from someone trying to spam their app.
Even may of the newly registered human accounts are starting to feel like spam bots.
This is why I see (well managed) government digital IDs as sensible moves. Apart from DDOS attacks, if bots have to “prove” who they are on each request it seems like a win-win.
If you want “papers, please” every time you back out of your driveway or go beyond your government-assigned oblast, then your suggestion is the digital version of the physical authoritarian nightmare that was imposed by totalitarianist regimes throughout history.
People have a right to complete anonymity, and should be able to go across the majority of the Internet just as they can go across most of the country.
That’s what you are missing.
Don’t get me wrong, I am also in favour of a single government ID, but in terms of combatting identity fraud, accessing public resources like single-payer healthcare, and making it easier for a person to prove their identity to authorities or employers.
It should not be used as a pass card for fundamental rights that normally would have zero government involvement.
You may be missing that it's easy and free for website owners to fix the problem. But it's hacker news after all. If somebody is bothered by a leaf falling on them on their walk to the corner store, the suggested solution here will be to have a full communist revolution.
In addition to pulling responses with huge amplification (40x, at least, for posting a single Facebook post to an empty audience), it's sending us traffic with fbclids in the mix. No idea why.
They're also sending tons of masked traffic from their ASN (and EC2), with a fully deceptive UserAgent.
The weirdest part though is that it's scraping mobile-app APIs associated with the site in high volume. We see a ton of other AI-training focused crawlers do this, but was surprised to see the sudden change in behavior on facebookexternalhit ... happened in the last week or so.
Everyone is nuts these days. Got DoSed by Amazonbot this month too. They refuse to tell me what happened, citing the competitive environment.
this is being disproved in the article posted:
>And so Anubis was enabled in the tar pit at difficulty 1 (lowest setting) when requests were pouring in 24/7. Before it was enabled, it was getting several hundred-thousand requests each day. As soon as Anubis became active in there, it decreased to about 11 requests after 24 hours, most just from curious humans.
apparently it does more than annoying users and making the site owner feel good (well, i suppose effective bot blocking would make the site owner quite good)
Normal and sane people understand this intuitively. If someone goes to a mechanic because their car is broken and the mechanic says "well, if you can tell that you car is broken, then you should be able to figure out how to fix it" - that mechanic would be universally hated and go out of business in months. Same thing for a customer complaining about a dish made for them in a restaurant, or a user pointing out a bug in a piece of software.
> Anubis uses a Proof-of-Work scheme in the vein of Hashcash
And if you look up Hashcash on Wikipedia you get https://en.wikipedia.org/wiki/Hashcash which explains how Hashcash works in a fairly straightforward manner (unlike most math pages).
I'll implement Anubis at low difficulty for all my projects and leave a decent llms.txt referenced in my sitemap and robots.txt so LLMs can still get relevant data for my site while.keeping bad bots out. I'm getting thousands of requests from China that have really increased costs, glad it seems the fix is rather easy.
It's even dumber than that, because by default anubis whitelists the curl user agent.
vsIn that case a better solution would be to take the site down altogether.
I love experimental data like this. So much better than gut reaction that was spammed when anubis was just introduced
The idea is to scare off bots and not normal humans.
Is the theory here that OpenAI, Anthropic, Gemini, xAI, Qwen, Z.ai etc are all either running bad scrapers via domestic proxies in Indonesia, or are buying data from companies that run those scrapers?
I want to know for sure. Who is paying for this activity? What does the marketplace for scraped data look like?
Why do you say that?
What is the point of these anti bot measures if organic HN traffic can nuke your site regardless? If this is about protecting information from being acquired by undesirable parties, then this site is currently operating in the most ideal way possible.
The information will eventually be ripped out. You cannot defeat an army with direct access to TSMC's wafer start budget and Microsoft's cloud infrastructure. I would find a different hill to die on. This is exactly like the cookie banners. No one is winning anything here. Publishing information to the public internet is a binary decision. If you need to control access, you do what Netflix and countless others have done. You can't have it both ways.
We need a better solution.
I think some of the folks running sites would rather have you go to the site and view the items “suggested based on your shopping history” (I consider these ads, the vendors might disagree), etc.
I’m more sympathetic to the people running sites than the LLM training scrapers, but these are two parties in a many-party game and neither one is perfectly aligned with users.
Very annoying. And you can't filter them because they look like legitimate trafic.
On a page with differents options (such as color, size, etc...) they'll try all the combinaisons, eating all the ressources.
On Safari or Orion it is merely extremely slow to load.
I definitely wouldn't use any of this on a site that you don't want delisted for cryptojacking.
> Here is a massive log file for some activity in the Data Export tar pit:
A bit of a privacy faux pas, no? Some visitors may be legitimate.
It also drives home that Anubis needs a time estimate for sites that don't use Anubis as a "can you run javascript" wall but as an actual proof of work mechanism that it purports to be its main mechanism
It shows a difficulty of "8" with "794 kilohashes per second", but what does that mean? I understand the 8 must be exponential (not literally that 8 hashes are expected to find 1 solution on average), but even as a power of 2, 2^8=256 I happen to know by heart, so thousands of hashes per second would then find an answer in a fraction of a second. Or if it's 8 bytes instead of bits, then you expect to find a solution after like 8 million hashes, which at ~800k is about ten seconds. There is no way to figure out how long the expected wait is even if you understand all the text on the page (which most people wouldn't) and know some shortcuts to do the mental math (how many people know small powers of 2 by heart)
https://web.archive.org/web/20260329052632/https://gladeart....
Should have have maybe prioritized differently...
If you have a logging stack, you can easily find crawler/bot patterns, then flag candidate IP subnets for blocking.
It's definitely whackamole though. We are experimenting with blocking based on risk databases, which run between $2k and $10k a year depending on provider. These map IP ranges to booleans like is_vpn, is_tor, etc, and also contain ASN information. Slightly suspicious crawling behavior or keyword flagging combined with a hit in that DB, and you have a high confidence block.
All this stuff is now easy to homeroll with claude. Before it would have been a major PITA.
One of the mistakes people assume is that AI capability means humanness. If you know exactly where to look, you can start to identify differences between improving frontier models and human cognition.
One concrete example from a forthcoming blog post of mine:
[begin]
In fact, CAPTCHAs can still be effective if you know where to look.
We ran 75 trials -- 388 total attempts -- benchmarking three frontier AI agents against reCAPTCHA v2 image challenges. We looked across two categories: static, where each image grid is an individual target, and cross-tile challenges, where an object spans multiple tiles.
On static challenges, the agents performed respectably. Claude Sonnet 4.5 solved 47%. Gemini 2.5 Pro: 56%. GPT-5: 23%.
On cross-tile challenges: Claude scored 0%. Gemini: 2%. GPT-5: 1%.
In contrast, humans find cross-tile challenges easier than static ones. If you spot one tile that matches the target, your visual system follows the object into adjacent tiles automatically.
Agents find them nearly impossible. They evaluate each tile independently, produce perfectly rectangular selections, and fail on partial occlusion and boundary-spanning objects. They process the grid as nine separate classification problems. Humans process it as one scene.
The challenges hardest for humans -- ambiguous static grids where the target is small or unclear -- are easiest for agents. The challenges easiest for humans -- follow the object across tiles -- are hardest for agents. The difficulty curves are inverted. Not because agents are dumb, but because the two systems solve the problem with fundamentally different architectures.
Faking an output means producing the right answer. Faking a process means reverse-engineering the computational dynamics of a biological brain and reproducing them in real time. The first problem can be reduced to a machine learning classifier. The second is an unsolved scientific problem.
The standard objection is that any test can be defeated with sufficient incentive. But fraudsters weren't the ones who built the visual neural networks that defeated text CAPTCHAs -- researchers were. And they aren't solving quantum computing to undermine cryptography. The cost of spoofing an iris scan is an engineering problem. The cost of reproducing human cognition is a scientific one. These are not the same category of difficulty.
[end]
I can't believe people are still using this as a generic anti-AI argument even though a decade ago people were insisting that there's no way AI can have the capabilities that frontier LLMs have today. Moreover it's unclear whether the gap even exists. Even if we take the claim that the grid pattern is some sort of fundamental constraint that AI models can't surpass, it doesn't seem too hard to work around by infilling the grids pattern and presenting the 9 images to LLMs as one image.
Your key finding is that humans process the grid as one visual scene — but that's a finding about sighted cognition.
Isn't this, like most things, a sensitivity specificity tradeoff?
How many real humans should be blocked from your system to keep the bots out?
What is the Blackstone ratio of accessibility?
Maybe I’m a bot, I gave up waiting before the progress bar was even 1% done.
JA4 fingerprinting works decently for the residential proxies.
Good luck banning yourself from the future.
> "The idea is that at individual scales the additional load is ignorable, ..."
Three minutes, one pixel of progress bar, 2 CPUs at 100%, load average 4.3 ...
The site is not protected by Anubis, it's blocked by it.
Closed.
If a webstie takes so long to verify me I'll bounce. That's it.
That’s how fast the landscape is changing.
And remember: while the report might have been released in 2024, it takes time to conduct research and publish. A good chunk of its data was likely from 2023 and earlier.
I don’t think the comments are bot-written, but it’s ironic that when I first clicked on this comment section 2 out of 3 comments (including the showdead one) were from someone trying to spam their app.
Even may of the newly registered human accounts are starting to feel like spam bots.
I may be missing something of course
People have a right to complete anonymity, and should be able to go across the majority of the Internet just as they can go across most of the country.
That’s what you are missing.
Don’t get me wrong, I am also in favour of a single government ID, but in terms of combatting identity fraud, accessing public resources like single-payer healthcare, and making it easier for a person to prove their identity to authorities or employers.
It should not be used as a pass card for fundamental rights that normally would have zero government involvement.