Gonna zag here. If you take a step back, cloudflare has been paving the path for pay for crawl. I think it's a noble and ambitious goal.
While I can understand why you would be alarmed, I can point to almost two decades of lamenting on this forum about how we need better ways of rewarding content creators than ads. Well, this is it.
Moreover, these products weren't built in a vacuum. Most threads about Anthropic and OpenAI have complaints about how these companies were built on stolen content. There's a reality we have to face here and we can't have it both ways.
Is it a better way though? The problem with ads has always been their abusive nature. From unrestrained pop ups and clickjacking in the 90s to today's pervasive surveillance and profiling.
Cloudflare turnstile is a pop up. This product is pervasive surveillance. It having a cf logo doesn't change that or ameliorate the many many abuses that two decades have shown are part and parcel.
You said "a similar range of services". Basic DDoS protection is probably important, and there are other companies doing that. Beyond that, it's less obvious.
They are advantageous to leverage in certain situations but essential they are not. We're used to, in the technology industry, looking for or creating problems to solve with services we are aware of. Moving back to necessity and need, do we really? Are we being objective? Most of the time, no.
Theres a few alternatives, but at a minimum yes you probably need their or a competitor's Name Servers and their public DNS. Rolling your own isn't very feasible.
The complaint is that the offer is a great deal with no downsides for consumers, and this is likely to result in Cloudflare having a lot of power (which they currently don't have) as a market maker. This position as market maker would grant them the power to extract economic rent from the web economy by charging both sides of the web provider and web consumer market to get access to the other.
For any one of their product there is a good opportunity to build an open source alternative or something like it! Can be hard to work around they have the benefit of being able to have negative unit economics on lots of infra products... But people succesfully built tons of alternatives to google analytics and similar.
Open source for bot protection specifically would be difficult. If I as a bot developer can see the tests you run I can just modify my bot to pass them (either trivially or by brute force).
It's really telling that they're starting to block anything automated ever that hasn't gone through their ID verification process. It's like literally every company that exists in the world is taking advantage of the dystopia at once
no, they're giving tools to their customers who can choose freely to block or not block bots. Without those tools, the people who run sites and offer content are just flying blind. I struggle to see how this is a bad thing in any way
Cool product launch, though it feels a little weird to me that Cloudflare sells agentic products alongside this new service that seems designed to block agentic usage of the web?
I expect there's much more going on than just mouse path detection but I can imagine that this is already tricky for touchscreens and for people using non-traditional mouse inputs (the thinkpad nub comes to mind - but it would also be bad optics to accidentally block people using accessibility mouse tools as bot users, though then this becomes a loophole for agentic browsing!)
In general though I think this is almost definitely a good thing to reduce agentic bot abuse & spam.
It’s less weird if you think there’s a difference between good bots and bad bots. They can provide services for good bots to use while helping people keep out the bad ones.
If a bot is simulating mouse movement but doing it badly then that’s a strong signal of shenanigans. A good bot will obey robots.txt and do nothing to hide that it’s a bot.
The problem is they have a strong incentive to consider any bot they get paid to host as a good bot, and they can bully competitors by declaring bots they host as bad bots
Arguments based on incentives can prove anything. We could just as easily say they have incentives to provide tools to both sides and let them fight it out.
It would be like Gmail automatically whitelisting email from other Gmail accounts or blacklisting email from competitors. Why should Google do that? Their customers are mostly strangers to each other and they want spam filters that work well.
> It would be like Gmail automatically whitelisting email from other Gmail accounts or blacklisting email from competitors. Why should Google do that? Their customers are mostly strangers to each other and they want spam filters that work well.
Crazy to post this when Google basically does do that with Gmail addresses. They may shove it into your “other” folder but they do absolutely nothing to stop free Gmail accounts from spamming and phishing. Gmail spam is the biggest threat by volume I’ve seen in years of managing email security for clients
> Cool product launch, though it feels a little weird to me that Cloudflare sells agentic products alongside this new service that seems designed to block agentic usage of the web?
Feels a little bit like the mob selling "protection" to shop keepers.
I think in 10/20 years from now, access to the Internet will be allowed only upon personal identification. Every website will be allowed to ask about your identity upon serving any content. Thats the only way I see this is going. The internet as it is right now does not have a future if majority of traffic will be done by agents. Thus, the cost of that traffic will have to be put on the users and since displaying ads doesnt make sense to agents, a paid access will be introduced (which is what cloudflare is slowly doing)
I implemented all of this in hCaptcha 6 years ago, not just to distinguish bot from human but also to recognize the keyboard/mouse behavior of the same person signing up for many accounts or testing multiple credit cards. This kind of abuse detection was a part of Cloudflare when they switched to hCaptcha in 2020 and I had thought they already implemented all this themselves four years ago when they transitioned away from hCaptcha in 2022.
I'm guessing it's going to lock the non-sighted//keyboard only users out of the anonymous Internet. I'm guessing if you log in and give up your anonymity they'll consider you not a bot.
I mean CF already forces 5 minutes of motorbike identification on anyone not in a whitelisted western country, so a small percentage of blind people is unlikely to worry them.
Yeah, you just have spinners that keep refreshing the page and spinning when you're using Firefox or a mobile device that's neither on Android nor iOS.
just like how smartphone developers, engineers locked people with no smartphone.Some of the users that can no longer get services are really old people, disabled. In some well developed places on earth, you can't even check in flights without a smartphone, it's not even possible to travel for them.
It seems like some of the main reasons people care so far are:
- Preventing automated credential stuffing
- Preventing bots from creating a bunch of fake accounts (eg free trial abuse, which can also lead to high twilio SMS bills!)
- Reducing payment fraud
- Blocking LLM scraping
- Blocking automated scalpers (!) eg for tickers or sneakers
I'm curious to see which use cases end up dominating as the reason companies care about this. And I'm hopeful that my agents will still have good ways for me to browse and do things on the web on my behalf - eg detect agents and route them to an agent path, rather than blocking them.
(I'm interested in tools for detecting AI agents and seeing how this shifts as bot traffic goes way up.)
Darwinium (darwinium.com) is another example. Approach here involves a combination of profiling and step-transition probabilities; idea is that a customer can ring-fence a particular area of a digital estate where they might want to challenge or block an agent - eg a payment, due to chargeback risks. Precursor at least for now seems more focused on site scraping multiple docs from the same site.
What prevents bots/agents from just adding "jitter" to their movements that mimics how humans move their cursor?
I know there are other signals being used but this one in particular seems like it wouldn't be hard to beat with a small amount of sophistication from the bot.
Beating this would require a large amount of sophistication, not a small amount.
Basic machine learning clustering will expose bots mouse+keyboard+touch behavior and discriminate them from humans.
It will also likely discriminate against anyone with a disability and therefore using affordances like eye tracking. Just imagine how different a person with only one hand would look compared to a “typical” user!! This shouldn’t be too much of a problem in the USA because no one is enforcing the ADA at the moment outside of California / Illinois / NY.
But I’m curious to hear from ‘eastdakota how they plan to guarantee that users with disabilities won’t be affected by these kinds of behavioral analysis. Cloudflare has such a massive footprint that it’s absolutely critical for them to err on the safe side of filtering, assuming they desire to be ethical.
The immoral thing for cloudflare to do would be to say “we just provide a ‘bot likeliness score’ and it’s up to each website to decide what threshold they need”. And then wave their hands and say “we’re not the ones blocking users with disabilities…the websites are the ones setting their thresholds too strictly”.
When you reach Cloudflare’s size … you own all the 2nd and 3rd order effects of your decisions.
This kind of data not only separates bots from humans - it’s pretty trivial to distinguish male vs female, right-handed vs left-handed, approximate age, native language (based on keyboard input patterns), state of injury (including tracking progression of healing), and a variety of different mental/physical disabilities. How one navigates a website tells you whether they are ADHD or schizophrenic or has Parkinson’s, and it can tell you about drug use/abuse: how well is this person’s Parkinson’s treatment working? What days of the week does that person tend to abuse amphetamines?
It is super difficult to mimic all of these signals in a way that would cluster the same as typical humans.
We used to say the same sorts of things about LLM prose, music, and image generation. Now just a few years later it can be very difficult to know for sure if something is made by AI or a human. There are still tells, but they are much more subtle and harder to spot, and models are still improving. Mimicing human mouse movement won't be any more of a challenge.
Humans are very inefficient when it comes to navigating the web, but also take actions pretty fast when completing forms. You don't really need advanced ML to see bots spend two seconds to read a full page, then spend 10 seconds just to click two buttons a human would click together in under 2 seconds. The amount of sophistication in bot detection peaks at about 'if user searches 20 queries in less than 5 minutes on our search engine and uses incognito, CAPTCHA them'.
Because of this, perfectly mimicking humans is not a good goal for a bot (as it is the case for AI in music), because they would become very inefficient, at least latency wise (throughput could be engineered around by scraping many unrelated webpages in parallel).
> It is super difficult to mimic all of these signals in a way that would cluster the same as typical humans.
Not really, beat ML with ML. I won't disclose how to do it, because who knows who might read this, but you can easily do it with a model trained for that purpose.
Sure - it's just hard for rando's to get tons and tons of real human interaction data to run a GAN against. "How to do the training" isn't the barrier for this, and not worth keeping a secret.
Like any other detection system you will always have determined adversaries that put in the work to bypass it.
But that doesn't mean you shouldn't still try to block the much larger number of less sophisticated/resourced adversaries that are using OOTB libraries and low-effort setups.
Sure, but of course since there's profit to be made defeating these systems once someone makes a program to defeat detection they'll sell it. Complicated attacks only stop simple attackers until a sophisticated attacker scripts & sells the exploit. Not that you shouldn't try, just don't expect defenses to last long-term.
are you sure it's non trivial? they posted a 2d image of what it looks like. a fairly simple model of the users wrist and mouse position doesn't seem crazy hard but the devil is in the details
In 2027 how many tokens will we spend to create the jitter, pre-jitter planning, post-jitter verification, and then cloudflare’s inevtiable counter-jitter
that's actually how you do it. adversarial systems like those are prime candidates. one agent develops detection mechanisms and the other agent defeats them. progression signal is easy to get.
and you bootstrap with existing javascript detection engines.
the challenge is usually the human input data, your objective is to be clustered among the humans and for that you need to know what humans look like.
this is not an open ended arms race, it will end once the bots approximate humans to a sufficient degree - false positive rate for detection will become unacceptable even if the detection system is slightly ahead.
Nothing. But you are already at a disadvantage because Cloudflare has seen far more real jitter data and you are up against that. It might work in the short term but after a while you start showing pretty obvious patterns. There's also a great variety of jitter data on specific websites or layouts that would be very easy to catch someone artificially emulating jitter
One interesting aspect is of course that the movement from the same user can be different depending on what type of mouse they use. I use a mouse at work on my PC, touchpad on my private laptop, and thinkpad nipple on work laptop. Three different profiles for one user.
Obviously different movements from a AI, but if we come to the day where mouse movement fingerprinting becomes another gatekeeper, there could be some interesting outliers.
There is nothing stopping a bot from moving their cursor like a human. This is basically just putting up a door with zero walls and telling people to stay out of your house.
All of these things are completely abusable/bypass-able and just annoying for actual humans who trigger flags.
> There is nothing stopping a bot from moving their cursor like a human.
Sure, we could write a library that slows the bot down and makes it move the cursor in procedurally-generated curves with a certain degree of noise added... but its all extra work, and it all slows the bots down. Presumably they wouldn't reveal that part of the secret sauce if it was all of the secret sauce
Acting like a human is something scapers already do. Using residential proxies, using latest Chrome user agents, not moving/typing as fast, etc. This is just 1 more layer, moving mouse naturally.
I have been noticing a lot of Cloudflare false positives where it keeps spinning on my sessions never actually redirecting me to the underlying page. If they keep just vibe coding and releasing a new solution every day, I am afraid it will be reflected in their services quality.
I can 1000% guarantee this will adversely impact assistive technology. You can tell it will because they don't mention any testing with regards to assistive technology.
Cloudflare has a lot of enterprise customers. Selling bot check to companies wanting to protect their content & also taking a cut out of payments for access by bots could be a good earner for them.
The examples of mouse movement, really reminds me what bot scripts looked like for Runescape back in 00s-10s. Early scripts were color-based and jumped the mouse around, and those were quickly caught. But over time, bot scripts developed into complex orchestrations; taking breaks, doing random actions spontaneously, moving the mouse naturally, logging sessions on different platforms (mobile, PC), even responding in chat.
There's been plenty of effort put into mimicking realistic / "human" behavior in writing video game bots, and every video game still has tons of bots despite the best efforts of the game devs.
You definitely can't win against bots - but you can definitely make the entire "game" (web at large, in this case) worse off for everyone else through this "always-online DRM" parallel.
It's a bleak world in terms of bots flooding the web, but out of all possible solutions, this seems to be preferable over invasive and identifying fingerprinting that everyone wants to roll out. Here's hoping that mouse movements aren't sufficiently unique as to be fingerprintable too.
I can’t wait for Cloudflare to decide my musical-rhythm enhanced typing and extraordinarily rapid and repeatable-pattern captcha clicking are somehow machine signifiers just like Google does: If I complete Google captchas at full speed it decides I’m a robot and challenges me endlessly, once 29 times in a row in two minutes or something. That’s what I get for having excellent spatial reflexes and mouse-clicker practice from Q3A sniping. So exhausted of the reversion to mediocre tendency of anti-bot systems, sigh.
You also screwed up by using Firefox. That's the #1 method Google uses to prove someone is a bot. If you are't participating in the Google panopticon, you are suspect.
I dislike bots as much as anyone else... when weird inquiries come through my company's lead form, it costs some time and attention to sort them.
But what makes Cloudflare so confident that automation always equates to "fraud and abuse?" If I send my agent to go retrieve some information, do they consider that fraud?
If I block various ad trackers does that trigger their "bot detection" incorrectly? Do I have any recourse? Or is Cloudflare appointing themselves judge, jury and executioner?
And let's not forget this little chestnut:
> 4. Privacy by design. Precursor was designed to collect signals that help to distinguish human patterns from automated and abusive patterns.
Ahh, so to "protect" against bots they're standing up a whole new regime of user surveillance and session-level monitoring. And they definitely won't be selling that, they promise. Got it.
This crap should be illegal. In the real world, I can authorize others to act on my behalf. The same should be true with software agents.
>keyboard activity, focus changes, and visibility. These events are serialized into a compact format and buffered in memory. At regular intervals, the buffered data is sent back to the evaluation layer for analysis.
I wonder how it'll handle those of us who try and use the mouse as infrequently as possible. I imagine the cognitive delay part would be largely telling. But it'll be interesting to see if I start getting blocked because I use vimium.
Yes, this is something almost all other anti-bot/fraud prevention solutions already does, and there are already bypasses developed simulating human mouse movement.
Not sure, but I struggle with skepticism for anyone who blocks archive.today, which cloudflare does, along with nextdns and others. Being blocked by such a large... apologies in advance for 'lack of better word' vernacular, cartel, is a near death sentence.
Although the blocking of archive.today goes back years, as can be verified through forum searches and archives with nextdns and others, I was not aware of this and have no excuse to dispute it. But for the record, the blocking predates 2026 by many years -- and my own records also verify this. That said, I think I need to learn more.
As a real user who uses an Ultimate Hacking Keyboard with the mouse layer, this frustrates me immensely. Yes I'm a corner case, but this is likely to make certain website not work for me because my lines are perfectly straight and my arcs zig-zag much like a bot might.
Considering the keyboard/mouse layer feels like an advancement to me, this feels like tech that will lock in the "old" way of doing things.
I really detest how adversarial the web is getting. I'm not a cloudflare hater but please, please consider people like me when rolling out stuff that affects millions or maybe even hundreds of millions or billions of people.
Precursor is a client-side, session-based verification system, built with privacy in mind, that uses dynamically injected JavaScript to continuously collect behavioral signals as visitors interact with your application.
So now instead of having the slow-axx Cloudflare turnstile slowing down your requests, you get surprised with a "You are a BOT!!!" while you are conducting your business on a website.
I already quickly close any website that I do not need for business purposes when it shows me the Cloudflare spinner. Now I might have to start considering competitors who do not implement this shit.
hmmm i think this is the first time i've seen a genuinely decent approach to blocking scraper/agents
that mouse cursor movement is very hard to replicate a real human with the amount of data that cloudflare has you could reproduce something close but cloudflare has seen probably trillions of movements that will be tough to beat
watching this carefully but i think this is the right approach
Yawn. Train a domain-specific model on human inputs and then run inference against that. At integration, you change what, one line of code with another? You at best raise the expense to bot, but in today's world, this isn't much compute expense. You can do it on 10-year-old Xenon processors, the same ones used by companies promoted on LowEndBox.
Skids already fall into the trap of using open source automation like playwright-extra-stealth.
Doesn’t seem healthy for the internet as a whole
While I can understand why you would be alarmed, I can point to almost two decades of lamenting on this forum about how we need better ways of rewarding content creators than ads. Well, this is it.
Moreover, these products weren't built in a vacuum. Most threads about Anthropic and OpenAI have complaints about how these companies were built on stolen content. There's a reality we have to face here and we can't have it both ways.
Cloudflare turnstile is a pop up. This product is pervasive surveillance. It having a cf logo doesn't change that or ameliorate the many many abuses that two decades have shown are part and parcel.
Until somewhere down the line. Like when half of Spain gets cut off due to an arbitrary block on a consolidated service facade...
We have antitrust regulations for such things.
They are advantageous to leverage in certain situations but essential they are not. We're used to, in the technology industry, looking for or creating problems to solve with services we are aware of. Moving back to necessity and need, do we really? Are we being objective? Most of the time, no.
I genuinely don't understand these generic complaint comments.
Are you complaining that they offer too much? Or do you believe nobody is offering similar services?
Just want to make sure I understand the real issue here, because that sounds like a lot of fearmongering to me.
where bots run rampant ?
trust me as an operator - I'm grateful Cloudflare exists.
I expect there's much more going on than just mouse path detection but I can imagine that this is already tricky for touchscreens and for people using non-traditional mouse inputs (the thinkpad nub comes to mind - but it would also be bad optics to accidentally block people using accessibility mouse tools as bot users, though then this becomes a loophole for agentic browsing!)
In general though I think this is almost definitely a good thing to reduce agentic bot abuse & spam.
If a bot is simulating mouse movement but doing it badly then that’s a strong signal of shenanigans. A good bot will obey robots.txt and do nothing to hide that it’s a bot.
Generally isn't a good bot one that respects robots.txt and is respectful of the site's resources by not being spammy?
It would be like Gmail automatically whitelisting email from other Gmail accounts or blacklisting email from competitors. Why should Google do that? Their customers are mostly strangers to each other and they want spam filters that work well.
Crazy to post this when Google basically does do that with Gmail addresses. They may shove it into your “other” folder but they do absolutely nothing to stop free Gmail accounts from spamming and phishing. Gmail spam is the biggest threat by volume I’ve seen in years of managing email security for clients
Feels a little bit like the mob selling "protection" to shop keepers.
Yeah so this mouse movement astrology is going to completely lock non-sighted/keyboard only users out of large swaths of the Internet isn't it.
Potato potato.
Yet all people are ok with it
Other products in the space:
- Foil (https://usefoil.com/), I'm biased, a friend is building this
- Kasada https://www.kasada.io/
- DataDome (https://datadome.co/)
- Castle (https://castle.io/)
- Fingerprint (https://fingerprint.com/)
- HUMAN (http://humansecurity.com/)
- Google Cloud Fraud Defense, which is basically the updated reCaptcha (https://cloud.google.com/security/products/fraud-defense?hl=...)
- this, Cloudflare Precursor
It seems like some of the main reasons people care so far are:
- Preventing automated credential stuffing
- Preventing bots from creating a bunch of fake accounts (eg free trial abuse, which can also lead to high twilio SMS bills!)
- Reducing payment fraud
- Blocking LLM scraping
- Blocking automated scalpers (!) eg for tickers or sneakers
I'm curious to see which use cases end up dominating as the reason companies care about this. And I'm hopeful that my agents will still have good ways for me to browse and do things on the web on my behalf - eg detect agents and route them to an agent path, rather than blocking them.
(I'm interested in tools for detecting AI agents and seeing how this shifts as bot traffic goes way up.)
I know there are other signals being used but this one in particular seems like it wouldn't be hard to beat with a small amount of sophistication from the bot.
Basic machine learning clustering will expose bots mouse+keyboard+touch behavior and discriminate them from humans.
It will also likely discriminate against anyone with a disability and therefore using affordances like eye tracking. Just imagine how different a person with only one hand would look compared to a “typical” user!! This shouldn’t be too much of a problem in the USA because no one is enforcing the ADA at the moment outside of California / Illinois / NY.
But I’m curious to hear from ‘eastdakota how they plan to guarantee that users with disabilities won’t be affected by these kinds of behavioral analysis. Cloudflare has such a massive footprint that it’s absolutely critical for them to err on the safe side of filtering, assuming they desire to be ethical.
The immoral thing for cloudflare to do would be to say “we just provide a ‘bot likeliness score’ and it’s up to each website to decide what threshold they need”. And then wave their hands and say “we’re not the ones blocking users with disabilities…the websites are the ones setting their thresholds too strictly”.
When you reach Cloudflare’s size … you own all the 2nd and 3rd order effects of your decisions.
This kind of data not only separates bots from humans - it’s pretty trivial to distinguish male vs female, right-handed vs left-handed, approximate age, native language (based on keyboard input patterns), state of injury (including tracking progression of healing), and a variety of different mental/physical disabilities. How one navigates a website tells you whether they are ADHD or schizophrenic or has Parkinson’s, and it can tell you about drug use/abuse: how well is this person’s Parkinson’s treatment working? What days of the week does that person tend to abuse amphetamines?
It is super difficult to mimic all of these signals in a way that would cluster the same as typical humans.
Because of this, perfectly mimicking humans is not a good goal for a bot (as it is the case for AI in music), because they would become very inefficient, at least latency wise (throughput could be engineered around by scraping many unrelated webpages in parallel).
Not really, beat ML with ML. I won't disclose how to do it, because who knows who might read this, but you can easily do it with a model trained for that purpose.
But that doesn't mean you shouldn't still try to block the much larger number of less sophisticated/resourced adversaries that are using OOTB libraries and low-effort setups.
and you bootstrap with existing javascript detection engines.
the challenge is usually the human input data, your objective is to be clustered among the humans and for that you need to know what humans look like.
this is not an open ended arms race, it will end once the bots approximate humans to a sufficient degree - false positive rate for detection will become unacceptable even if the detection system is slightly ahead.
I'm sure, they can add a jitter, but then you just change how you detect / weight detection.
Obviously different movements from a AI, but if we come to the day where mouse movement fingerprinting becomes another gatekeeper, there could be some interesting outliers.
All of these things are completely abusable/bypass-able and just annoying for actual humans who trigger flags.
Sure, we could write a library that slows the bot down and makes it move the cursor in procedurally-generated curves with a certain degree of noise added... but its all extra work, and it all slows the bots down. Presumably they wouldn't reveal that part of the secret sauce if it was all of the secret sauce
I'm not sure if I'm talking about the scrapers or Cloudflare at this point. Probably both. Probably the same pool of forced laborers.
There's been plenty of effort put into mimicking realistic / "human" behavior in writing video game bots, and every video game still has tons of bots despite the best efforts of the game devs.
You definitely can't win against bots - but you can definitely make the entire "game" (web at large, in this case) worse off for everyone else through this "always-online DRM" parallel.
But what makes Cloudflare so confident that automation always equates to "fraud and abuse?" If I send my agent to go retrieve some information, do they consider that fraud?
If I block various ad trackers does that trigger their "bot detection" incorrectly? Do I have any recourse? Or is Cloudflare appointing themselves judge, jury and executioner?
And let's not forget this little chestnut: > 4. Privacy by design. Precursor was designed to collect signals that help to distinguish human patterns from automated and abusive patterns.
Ahh, so to "protect" against bots they're standing up a whole new regime of user surveillance and session-level monitoring. And they definitely won't be selling that, they promise. Got it.
This crap should be illegal. In the real world, I can authorize others to act on my behalf. The same should be true with software agents.
So it's a keylogger?
I expect this will be effective for maybe a day.
Not a fan
Considering the keyboard/mouse layer feels like an advancement to me, this feels like tech that will lock in the "old" way of doing things.
I really detest how adversarial the web is getting. I'm not a cloudflare hater but please, please consider people like me when rolling out stuff that affects millions or maybe even hundreds of millions or billions of people.
I already quickly close any website that I do not need for business purposes when it shows me the Cloudflare spinner. Now I might have to start considering competitors who do not implement this shit.
that mouse cursor movement is very hard to replicate a real human with the amount of data that cloudflare has you could reproduce something close but cloudflare has seen probably trillions of movements that will be tough to beat
watching this carefully but i think this is the right approach
Skids already fall into the trap of using open source automation like playwright-extra-stealth.