Fixing what YouTube couldn’t.

[LamerGamerYT]

4 minutes ago, KendallDoesCoding said:

Yup, I am sure he can improve the code, no say about that. If, your interested in contributing, be sure to check out: https://github.com/ThioJoe/YT-Spammer-Purge

The best bet would be to link the code with an obfuscated native binary file. I would love to help out; however, I don't have the time due to personal and commercial projects.

[KendallDoesCoding]

3 minutes ago, LamerGamerYT said:

The best bet would be to link the code with an obfuscated native binary file. I would love to help out; however, I don't have the time due to personal and commercial projects.

Yes. Ok cool!

[creesch]

45 minutes ago, aurel.dorel0 said:

It bothers me to see how well this video is being received, when there is so much misinformation/oversimplification in it.

 

This 3rd party filter isn't "better" than youtube's own, and doesn't do anything that the youtube filter can't. It's just that bots have an incentive to break the youtube filter, whereas they have no such incentive for a random 3rd party one, which only a minority of people will use.

 

Here's how the spam-fighting process works: you come up with a filter, implement it on your website, then bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters. Then that one pattern that was found to bypass your filters gets copy-pasted into every comment, until you have to add a new filter for it; and the cycle continues.

 

The only, and I repeat, only reason why this filter is "better" than youtube, is specifically because it's not implemented into youtube (and as such, hasn't had an incentive to be broken). if it were implemented into youtube, it would be broken in less than a few days. As a side note, you only see this filter run among the thousands of comments which bypassed youtube's (current) filters, completely disregarding the millions which were already blocked by it. You're not comparing filter A and filter B on a random sample to see which has better detection ratio, you're just running filter B on whatever bypassed filter A. That tells you nothing.

 

There are so many nuances about fighting spam and developing proper filters and techniques, it really bothers me to see this massive oversimplification get so much positive feedback. For example, it's very often beneficial to leave spam comments up for a few minutes and only delete them later. Why? Because as previously mentioned, "bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters". So if you let some comments randomly slip through (and only delete them later), you might trick some bots into believing their comments are not being caught by your filters, when in fact they are.

 

And none of this is being addressed in this video. It's just "look at how amazing this 3rd party filtering system is, which has never been attacked by real bots and has not been properly compared against youtube's defenses!". I don't know, I just got really triggered by this video.

Eh, your comment comes awfully close to just simplifying it to binary terms and taking position in one end of the spectrum and digging your heels in. There is a whole lot more to it and rather more complex as you make it out to be. 

For starters, some very specific patterns that have been common for the past months and for some reason have not been picked up by YouTube for whatever reason. So at a basic level we can already verify that it isn't about YouTube not being able to, but rather YouTube simply not doing it. Which brings me to an additional point, it can also simply be the case that these specific sorts of comments have not negatively impacted whatever KPIs YouTube PMs are using to justify what should have focus. Those KPIs aren't necessarily the sort of thing that are about a good user experience but rather what impacts revenue. Comments always have been under prioritized by YouTube to begin with, if they had valued discussion more under videos the entire commenting section would be structured differently to begin with. 

Which also easily translates to dealing with spam, because YouTube has different priorities it is entirely possible for a community project to do much better than YouTube itself. 

 

This isn't unique to YouTube as a platform, other platforms like reddit also heavily depend on third party community tooling as the company itself simply was not after the same metrics. See also my comment earlier in this thread. 

 

Yes it is entirely possible that a community maintained filter also will be figured out by spammers. But at the same time the community can also much better anticipate on it as they are effectively boots on the ground experiencing the situation first hand. I am not saying that this tool specifically is the holy grail, but it is a damn good start in improving how channels can manage their comment section.  

Here is the thing, you aren't even entirely wrong. Spammers hate figuring out specific patterns for individual communities, which is why community unique filters and approaches are a wonderful thing. As I said in the earlier comment, reddit has had automoderator for years now, which has been a great asset in this regard. Also not perfect but I was amazed to find that besides this tool there isn't widespread tooling like that available for youtube channels who value their comment section. 

 

 

Edit: 

 

To be clear, when I say that I am amazed at how spartan the youtube comment moderation situation is I really mean that I am absolutely baffled. It makes me wonder if API support by google is just so annoying to deal with or if there are other things at play. 

 

I mentioned reddit as a platform with a lot of awesome third party open source projects that aid in moderation, but Discord is also a great example as a platform where you can choose from a myriad of moderator bots that you either host yourself or rent as basically a cloud service. Most of these bots offer all sorts of filtering mechanisms (often inspired on automoderator) and sophisticated spam fighting tools. 

 

So yeah, it most certainly is possible for the community to make better tools and filters than the big corporation. For the simple reason that the community cares more, has ears on the ground and has the incentive to actually tackle it. The only reasons I can think of that this hasn't been figured out for youtube yet is the fact that the google API situation always has been tricky and the fact that channels themselves are more insular and aimed at content first and somewhat less the discussion below it.  But, like I said in my earlier comment, it is awesome they are slowly figuring this out

 

[Ultraforce]

1 hour ago, Ashley MLP Fangirl said:

that moment when linus didn't know how to use github 2 months ago and is now comfortably using it to download and run a commandline only tool...

I imagine it helps that Anthony is the writer of the video as the writers usually show/set up the demo for Linus, I believe. Which is why there were GitHub stuff before the whole Linux Challenge. Also, probably helps Linus that there are releases rather than a project which cloning is preferred.

[Azzychan]

53 minutes ago, CarlEd77 said:

I love the fact that we can report the spammers, even when it isn't our channel.

Community action - that's how it should be.

Yes, we can always do that.  Thing is that it gets annoying when you get them on all comments.

I hardly comment on any video because of the amt. of porn spam I get replying to my comments.

[ThioJoe]

11 minutes ago, creesch said:

Eh, your comment comes awfully close to just simplifying it to binary terms and taking position in one end of the spectrum and digging your heels in. There is a whole lot more to it and rather more complex as you make it out to be. 

For starters, some very specific patterns that have been common for the past months and for some reason have not been picked up by YouTube for whatever reason. So at a basic level we can already verify that it isn't about YouTube not being able to, but rather YouTube simply not doing it. Which brings me to an additional point, it can also simply be the case that these specific sorts of comments have not negatively impacted whatever KPIs YouTube PMs are using to justify what should have focus. Those KPIs aren't necessarily the sort of thing that are about a good user experience but rather what impacts revenue. Comments always have been under prioritized by YouTube to begin with, if they had valued discussion more under videos the entire commenting section would be structured differently to begin with. 

Which also easily translates to dealing with spam, because YouTube has different priorities it is entirely possible for a community project to do much better than YouTube itself. 

 

This isn't unique to YouTube as a platform, other platforms like reddit also heavily depend on third party community tooling as the company itself simply was not after the same metrics. See also my comment earlier in this thread. 

 

Yes it is entirely possible that a community maintained filter also will be figured out by spammers. But at the same time the community can also much better anticipate on it as they are effectively boots on the ground experiencing the situation first hand. I am not saying that this tool specifically is the holy grail, but it is a damn good start in improving how channels can manage their comment section.  

Here is the thing, you aren't even entirely wrong. Spammers hate figuring out specific patterns for individual communities, which is why community unique filters and approaches are a wonderful thing. As I said in the earlier comment, reddit has had automoderator for years now, which has been a great asset in this regard. Also not perfect but I was amazed to find that besides this tool there isn't widespread tooling like that available for youtube channels who value their comment section. 

 

 

Yea pretty much. It's in no way a be-all-end-all solution. There are definitely patterns spammers have been using for literally years that are so INCREDIBLY obvious that I can only conclude YouTube literally does not care at all. Those are the ones that the filter targets first.

 

Of course the spammers will try to get around it, and will soon enough. My thinking is not to stop them, but hopefully force them to use more and more convoluted means to get their message across, so that it makes their scams more obvious and less effective, therefore less profitable and less worth it.

 

Also I added multiple methods for searching for this exact reason. If you start noticing spammers using some technique to get around, you can just search for that themselves with one of the other modes.

 

 

Quote

LamerGamerYT:

The obfuscation technique used by ThioJoe for the obfuscation of regex patterns is awfully basic. The obfuscation methods used need to be improved upon; else, there is no point as bots will keep evolving.

I use the term 'obfuscated' really loosely. None of the code is obfuscated, only a few hardcoded keywords. I know the scammers that are at the top of the pyramid are going to figure it out anyway, so it's basically just to prevent the dumbest scammers (who just buy turn-key scam packs basically) from being able to quickly scan the code and look at the exact strings.

[Nystemy]

1 hour ago, aurel.dorel0 said:

It bothers me to see how well this video is being received, when there is so much misinformation/oversimplification in it.

 

This 3rd party filter isn't "better" than youtube's own, and doesn't do anything that the youtube filter can't. It's just that bots have an incentive to break the youtube filter, whereas they have no such incentive for a random 3rd party one, which only a minority of people will use.

 

Here's how the spam-fighting process works: you come up with a filter, implement it on your website, then bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters. Then that one pattern that was found to bypass your filters gets copy-pasted into every comment, until you have to add a new filter for it; and the cycle continues.

 

The only, and I repeat, only reason why this filter is "better" than youtube, is specifically because it's not implemented into youtube (and as such, hasn't had an incentive to be broken). if it were implemented into youtube, it would be broken in less than a few days. As a side note, you only see this filter run among the thousands of comments which bypassed youtube's (current) filters, completely disregarding the millions which were already blocked by it. You're not comparing filter A and filter B on a random sample to see which has better detection ratio, you're just running filter B on whatever bypassed filter A. That tells you nothing.

 

There are so many nuances about fighting spam and developing proper filters and techniques, it really bothers me to see this massive oversimplification get so much positive feedback. For example, it's very often beneficial to leave spam comments up for a few minutes and only delete them later. Why? Because as previously mentioned, "bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters". So if you let some comments randomly slip through (and only delete them later), you might trick some bots into believing their comments are not being caught by your filters, when in fact they are.

 

And none of this is being addressed in this video. It's just "look at how amazing this 3rd party filtering system is, which has never been attacked by real bots and has not been properly compared against youtube's defenses!". I don't know, I just got really triggered by this video.

Spam filtering is indeed a nuanced battlefield that is also a very hard fight to win.

 

Google/Youtube could take the next step of blocking IP addresses that the spammers use, but that will only make ALL VPNs stops working and people complain that Google blocks VPNs, Ie, blocking IP addresses isn't a tool one can use without major negative feedback.

 

Another solution to calming down the fight is to restrict API access to validated users. _{(Similar to how Youtube already requires a new channel to gain a minimum amount of subscribers, have a minimum annual view duration and also have 0 community strikes for it to be eligible to have advertisements. One can't just throw up a channel and make money from the first view from some random video one copied from someone else.)}

 

But even with API access behind validation bots could still just click about the site like Joe average. Here one can look at the rate at which a user performs actions, like no human posts 100's of comments a second. But limiting the speed down to realistic levels is only going to lead to more spam accounts instead twiddling at the problem in parallel. And potentially with more variation.

 

In the end.

Spam filtering is as you say not simple.

"Easiest" way I can think of to filter spam is to just have a team of developers individually make their own filters. Then just use all of them together to filter out content. Likely by giving content a score based on how many filters a comment got flagged by. If it flagged by only a select few it likely is fine, if flagged by a lot it likely is spam.

An advantage of this multi filter approach is that one can roll out a new filter to the group quite often without this new filter needing to be "amazing" at the job since it isn't alone. _{(for example, if one has an 24 hour filter release schedule and 10 developers, then they each ave 57 hours (7 work days) to develop their filter before making a new one. Assuming they only work 8 hours a day 5 days a week. (400 man hours and 7 new filters a week))}

 

Downside is that with sufficiently many parallel filters, a regular comment is likely going to be considered bad by some of them. But as stated, one only uses the filters to score the comment. And different filters might work better or worse for different channels. So together with corrections made by the channel moderators one can "train" the filter scoring for that channel. _{(ie, if a moderator states that "this comment is fine", then the filters flagging it will have a lesser effect for future comments on that channel. And vice versa for when a moderator flags a comment as spam.)}

 

Another downside is that one would need to send a comment through the many filters. And this might take a bit of computational effort. So over time one might need to trim out the "redundant" filters. Or decide on what one regards as the primary filters for a channel. Like only use filters with a high score for that channel. _{(Here one might re-filter a comment with "all" filters when a moderator makes a decision as to train the model a bit more incase some mistakenly low scored filter is actually useful, or if spam behavior has changed.)}

[creesch]

7 minutes ago, ThioJoe said:

Of course the spammers will try to get around it, and will soon enough. My thinking is not to stop them, but hopefully force them to use more and more convoluted means to get their message across, so that it makes their scams more obvious and less effective, therefore less profitable and less worth it.

 

Yeah makes sense. I am not sure if you saw my edit btw but can I suggest you also look for inspiration on other platforms? I mentioned reddit as I just happen to have a ton of domain knowledge there given that I have been involved there a lot over the years as one of the creators of the moderator toolbox for reddit browser extension which is in use by over 20k moderators on reddit (not mentioning it as a brag but to give some context to where I am coming from). However in the context of youtube channels and the comment section I am more thinking about automoderator which basically is a filtering bot that is incredibly flexible in setting up all sorts of filters. Having that sort of tooling available for content creators would allow them to tailor their filters for what they see and basically makes it random enough that spammers will have a hard time figuring out how to deal with them. The awesome thing about automoderator is that even though it is now integrated into reddit it actually used to be a third party bot that someone made which is why I think it is one of the best examples of giving communities the power to tackle these things themselves. 

 

In addition to that there is also the huge eco system of discord moderation bots offering all sorts of clever ways to detect spam, deal with spam and prevent spam. 

 

Also, if it is not you who looks to these tools for inspiration (because I can understand you already being busy enough with the tool you made) I do hope other people within the YouTube creator community do take note and see the possibilities hopefully kicking of a similar ecosystem of comment moderation tools allowing creators to generally have a better handle on the comment section. 

[rcmaehl]

1 hour ago, LamerGamerYT said:

I agree that it does not prevent spammers; instead, it makes it easier to delete their comments. The program, however, will make it easier to delete their comments only for a finite amount of time. That limited time reduces steeply if even a regular person like me can de-obfuscate the regex expressions used in less than five minutes.

 

All I'm saying is ThioJoe should improve upon the obfuscation techniques used.

I've brought this up a couple times now. Once on github and once elsewhere on the forums. I wasn't sure if it was intentional or not when I first saw it. There's quite a few tools available to help with quick and dirty obfuscation such as:

• imports
• parameters by reference
• eval()
• local variables with the same name as global variables

[Ne0Bonde]

I will try to build an app for Mac and iOS on the basis of source code, hope that helps.

[JH17002723]

Am I like Stupid or something?
Where do i find the .exe he is writing about in the github
@ThioJoe

Edited February 1, 2022 by JH17002723
Tag, specified

[orbitalbuzzsaw]

This is actually really impressive, obviously not perfect but it is limited by Youtube fuckery. Kudos to ThioJoe.

[Rex Hite]

In a way, this is kind of adblocking.

[BuckGup]

Damn LTT should not have made a video on this. The code is messy, has tons of deprecated features still laying around, bad comments, could be much more efficient, and the "hidden" method he uses is very basic and already bypassed. So really all that has been done is congesting Youtube's reporting system and creating a false perception of combating the problem. 

 

Alphabet has infinitely more resources than ThioJoe and access to the source code of Youtube. So I would have much rather seen a PSA style of video calling upon Youtube to take action rather than a script kiddie creating headaches for Youtube devs.

[Olgyd]

  

5 hours ago, aurel.dorel0 said:

It bothers me to see how well this video is being received, when there is so much misinformation/oversimplification in it.

 

This 3rd party filter isn't "better" than youtube's own, and doesn't do anything that the youtube filter can't. It's just that bots have an incentive to break the youtube filter, whereas they have no such incentive for a random 3rd party one, which only a minority of people will use.

 

Here's how the spam-fighting process works: you come up with a filter, implement it on your website, then bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters. Then that one pattern that was found to bypass your filters gets copy-pasted into every comment, until you have to add a new filter for it; and the cycle continues.

 

The only, and I repeat, only reason why this filter is "better" than youtube, is specifically because it's not implemented into youtube (and as such, hasn't had an incentive to be broken). if it were implemented into youtube, it would be broken in less than a few days. As a side note, you only see this filter run among the thousands of comments which bypassed youtube's (current) filters, completely disregarding the millions which were already blocked by it. You're not comparing filter A and filter B on a random sample to see which has better detection ratio, you're just running filter B on whatever bypassed filter A. That tells you nothing.

 

There are so many nuances about fighting spam and developing proper filters and techniques, it really bothers me to see this massive oversimplification get so much positive feedback. For example, it's very often beneficial to leave spam comments up for a few minutes and only delete them later. Why? Because as previously mentioned, "bots will post literally tens of thousands of comments with different patterns until ONE of them bypasses your filters". So if you let some comments randomly slip through (and only delete them later), you might trick some bots into believing their comments are not being caught by your filters, when in fact they are.

 

And none of this is being addressed in this video. It's just "look at how amazing this 3rd party filtering system is, which has never been attacked by real bots and has not been properly compared against youtube's defenses!". I don't know, I just got really triggered by this video.

  

1 hour ago, BuckGup said:

Damn LTT should not have made a video on this. The code is messy, has tons of deprecated features still laying around, bad comments, could be much more efficient, and the "hidden" method he uses is very basic and already bypassed. So really all that has been done is congesting Youtube's reporting system and creating a false perception of combating the problem. 

 

Alphabet has infinitely more resources than ThioJoe and access to the source code of Youtube. So I would have much rather seen a PSA style of video calling upon Youtube to take action rather than a script kiddie creating headaches for Youtube devs.

 

ThioJoe himself explained his motives better in his own video, but tl;dr the tool is less to combat spam / make a better filter than YouTube but rather to automate comment deletion which YT's UI doesn't allow (you have to delete them one by one).

[KiseloMlqko]

Hello, 

So after watching the video, finnaly i have a place to discuss that same problem but the other way around. 

The problem is that Youtube deletes most of the comments under my videos and not even moving them for review, just straight up delete. My guess was that its a spam filter, but there was no way getting back the comments. I can see that there was 15 comments but just 3 were there. 

[nideii]

Generally speaking, filtering comment is very hard, and for state of the art method is very very computational heavy, and not even close to perfect.

 

For ThioJoe's script, it was only good for certain types of comments, meaning ThioJoe needs to update the algorithm if spammers got a new way to spam.

 

What's more, comment spamming is not a huge of a deal, simply because it would not drive users away from the platform.

 

To be honest, if all creators put a "please report spam in the comments section" screen at the end of the video would be way more effective than any softwares, but I bet nobody would waste the end screen for cleaning comments. (Only a small number of users take action will be enough, imo)

 

YouTube creators care more about subscribers or Clickthrough rate than people commenting on the video. So if dose not hurt making money, it would always be "nice to have".

[ThioJoe]

1 hour ago, nideii said:

Generally speaking, filtering comment is very hard, and for state of the art method is very very computational heavy, and not even close to perfect.

 

For ThioJoe's script, it was only good for certain types of comments, meaning ThioJoe needs to update the algorithm if spammers got a new way to spam.

 

What's more, comment spamming is not a huge of a deal, simply because it would not drive users away from the platform.

 

To be honest, if all creators put a "please report spam in the comments section" screen at the end of the video would be way more effective than any softwares, but I bet nobody would waste the end screen for cleaning comments. (Only a small number of users take action will be enough, imo)

 

YouTube creators care more about subscribers or Clickthrough rate than people commenting on the video. So if dose not hurt making money, it would always be "nice to have".

You might think differently if it was your face scammers were using to literally steal money from people directly below a video you worked hard on. The porno bots are one thing, they're a nuisance. The impersonators though, make my blood boil, and are the main reason I wrote the script in the first place.

[Naveen Prashanth]

11 hours ago, James said:

YouTube’s spam problem has gotten out of hand… And it’s up to the community once again to do what YouTube can’t. Thankfully, the community has delivered…

 

https://github.com/ThioJoe/YT-Spammer-Purge

https://www.youtube.com/watch?v=2tRppXW_aKo

https://www.youtube.com/watch?v=c6ebWvay8dE

 

See this nonsense spamming in the spammer-against video ...I'm talking about the one who replied to me in the comments 

 

[RejZoR]

Boy I can't wait for Youtube to get DDOS-ed by this tool and pull the API support for it... This won't work because of nature how it brute forces through all comments, meaning you're basically harassing their servers with it. One doing it is not an issue. 5 million people doing it is.

 

It is kinda hilarious how Google always brags about Ai this and Ai that and yet all their expertise and technologies around magical Ai can't deal with mostly super obvious spam. Gotta give Wordpress a credit because the amount of crap they filter out of their comment sections is insane. I've been using their platform for my blog for many years and the amount of crap that gets stuck in the net and not being visualized on the blog is amazing. And in all this time I maybe had to moderate like 10 messages myself because they posted a lot of links in a single comment as a reply to my post.

[creesch]

9 hours ago, BuckGup said:

Damn LTT should not have made a video on this. The code is messy, has tons of deprecated features still laying around, bad comments, could be much more efficient, and the "hidden" method he uses is very basic and already bypassed. So really all that has been done is congesting Youtube's reporting system and creating a false perception of combating the problem. 

 

Alphabet has infinitely more resources than ThioJoe and access to the source code of Youtube. So I would have much rather seen a PSA style of video calling upon Youtube to take action rather than a script kiddie creating headaches for Youtube devs.

Sorry, but this is a completely ridiculous take you are writing down there. For starters, code doesn't have to be perfect to work. Good clean code helps in maintaining a software product that is for sure, it is however not needed to make a good working product. 

 

Secondly, Alphabet might have a lot of resources but as I already pointed out elsewhere if they decide not to allocate them to this issue then that is not relevant. Then there is also the fact that the community generally speaking has better insights than developers from Alphabet tackling the problem from the outside. So even with limited resources, it is entirely possible that people from within the community come up with tools fixing issues that Alphabet isn't tackling. 

 

I am not going to repeat entirely what I wrote elsewhere, but see also this comment and this comment where I get into this deeper and also provide actual examples of the community stepping up and providing working solutions on other platforms. 

 

 

 

6 hours ago, nideii said:

 

For ThioJoe's script, it was only good for certain types of comments, meaning ThioJoe needs to update the algorithm if spammers got a new way to spam.

 

 

I don't see an issue with that, in fact it is how most spam fighting and general moderation of comment sections and forums works in general. See also the two comments Iinked in reply to BuckGup above where I go into community made tools on other platforms that allow individual communities to do exactly that with great succes. 

 

Quote

What's more, comment spamming is not a huge of a deal, simply because it would not drive users away from the platform.

A so you are the product manager at YouTube who manages the spam fighting tools! Thank you for this insight in one of the KPIs you are using to decide if this is an issue. 

You are describing exactly why it might be that YouTube will not fix it, that doesn't however mean it isn't a problem for the community. Yes people are not being driven away from the platform, but they are discouraged from actually engaging in the community. For YouTube creators who care about the community and discussion there this is an actual issue, so it makes sense that the community then will step in and do it themselves. 

[creesch]

1 hour ago, RejZoR said:

Boy I can't wait for Youtube to get DDOS-ed by this tool and pull the API support for it... This won't work because of nature how it brute forces through all comments, meaning you're basically harassing their servers with it. One doing it is not an issue. 5 million people doing it is.

 

It is kinda hilarious how Google always brags about Ai this and Ai that and yet all their expertise and technologies around magical Ai can't deal with mostly super obvious spam. Gotta give Wordpress a credit because the amount of crap they filter out of their comment sections is insane. I've been using their platform for my blog for many years and the amount of crap that gets stuck in the net and not being visualized on the blog is amazing. And in all this time I maybe had to moderate like 10 messages myself because they posted a lot of links in a single comment as a reply to my post.

 

There are a few ways this can go, other platforms actually embraced these sorts of tools in the past or begrudgingly integrated them. A great example I mentioned a few times now is automoderator on reddit which started out as a third party bot on reddit. Eventually it got special permissions with higher rate limits as it filled a huge need reddit couldn't solve themselves at the time. Eventually the creator of that bot was hired by reddit as a developer and basically used that as an opportunity to integrate it into reddit itself. 

 

As far as google Ai goes, they have very capable algorithms and machine learning tools. However, it is also very clear that they are very picky where they fully apply them and tune them to perfection. It pretty much gives you a clear indication of what their internal priorities are and aren't. It's clear that for mail they have gone full in, likely because that's where it will actually negatively impact certain  KPIs if too much spam gets through. Clearly for YouTube comments they have gone for a "good enough" approach and what is leaking through now hasn't caused them to revisit their approach yet as it hasn't triggered any of the KPIs they deem valuable. Something similar I suspect is going on with the YouTube algorithm itself, it really falls short in showing actually new and interesting content other than a rehash of stuff you basically already know. But, it probably works well enough for the majority of people or at least in the sense that it satisfies their KPIs for click through, watch time, loiter time on YouTube, etc that they simply don't feel the need to actually make it as sophisticated as some people would like them to. 

[BuckGup]

15 hours ago, Olgyd said:

  

  

 

ThioJoe himself explained his motives better in his own video, but tl;dr the tool is less to combat spam / make a better filter than YouTube but rather to automate comment deletion which YT's UI doesn't allow (you have to delete them one by one).

Sure I get that but he's making more of a headache for youtube as he's automated a tool that wasn't automated. It could simply be for no reason at all or maybe it was designed that way to prevent congestion as Youtube is relying on automated bot detection and provides manual control to users to catch what was missed. With that said it still doesn't excuse Youtube for how shitty it's bot detection is. There's also the scenario of the "bots" not actually being a bot but a human. Bot farms are very lucrative so it's not surprising if they hire cheap labor to post things that are more hidden or less bot like. As they will have better outreach and conversions as the posts are less likely to be taken down or flagged as spam. So the the extra labor and time costs are outweighed compared to fully automating it

[BuckGup]

7 hours ago, creesch said:

For starters, code doesn't have to be perfect to work. Good clean code helps in maintaining a software product that is for sure, it is however not needed to make a good working product. 

This is contradictory as to classify something as "a good working product" would mean it doesn't have random code laying around, bugs, poor comments, and lack of efficient design. There could be a massive exploit in ThioJoes code due to deprecated features hanging around or simply overlooked as the code is hard to read. I'm not saying what ThioJoe is trying to do is bad or shouldn't be talked about but his execution has royally shot himself in the foot. He could have waited a month longer and written test cases for his code, had a private code review, asked the community to begin with to see what the most effective solution is and architect off that, or simply asking Youtube themselves. Because currently with what I see I don't think this will be effective for much longer unless major changes happen and Youtube doesn't combat it.

[creesch]

5 minutes ago, BuckGup said:

This is contradictory as to classify something as "a good working product" would mean it doesn't have random code laying around, bugs, poor comments, and lack of efficient design.

You are conflating a solid and maintainable code base with actual functionality. You can have a product that performs exactly according to specifications, but is still a mess underneath. Even if you argue that it could work better or isn't perfect, that might be entirely true, but perfect and good are still not the same thing and in this case more than adequate is still infinitely better than the alternative which is nothing. You are also very much focussing on just one aspect of the bigger argument, making it seem like a very one-sided thing, which it isn't. So if you'd like to continue to discuss the subject I'd be happy to, but before that I would like to invite you to actually respond to the entirety of what I said.