« first day (1475 days earlier)      last day (2349 days later) » 

12:12 AM
[ SmokeDetector | MS ] Repeating characters in title: Hihhhhhhhhhhhhhhh by Alexander Brown on stackoverflow.com
 
12:56 AM
Yo @ArtOfCode you around? I want to talk to you about something related to more-than-3 flags autoflagging
I might have mentioned this before but now that I'm writing code for something related I wanted to bring it up
 
semi-around
 
OK so something I'm making currently for Metasmoke is a thing that takes in a target accuracy and outputs a pretty much optimal flagging condition with at least that accuracy
 
[ SmokeDetector | MS ] Potentially bad keyword in body: Adding slides to an existing carousel manually by Ian G. on wordpress.SE
 
The implementation is not important but the implications are. Originally I was just thinking it would run once a day and then the optimal condition du jour for 98.5% would show up on the dashboard or whatever
 
1:00 AM
API quota rolled over with 7650 requests remaining. Current quota: 19999.
math: 274
stackoverflow: 175
superuser: 43
physics: 39
english: 26
electronics: 22
askubuntu: 20
serverfault: 19
apple: 19
unix: 17
codegolf: 17
bitcoin: 16
drupal: 15
es.stackoverflow: 14
security: 14
academia: 13
meta: 12
worldbuilding: 10
scifi: 10
dba: 9
ell: 8
webapps: 8
gaming: 8
softwareengineering: 8
mathoverflow.net: 8
diy: 7
chemistry: 7
ethereum: 7
rpg: 7
money: 7
salesforce: 7
stats: 6
android: 6
wordpress: 6
graphicdesign: 6
history: 6
arduino: 6
civicrm: 6
engineering: 5
gis: 5
tex: 5
 
but I kinda have a better idea
what if we ditch having people manually enter a reason weight/number of reasons/rep condition and instead
 
they put desired accuracies and every day the accuracies are turned into the real flagging conditions
 
@quartata define optimal?
 
@ArtOfCode number of TPs caught
 
1:03 AM
Needs to factor in FPs somehow
Not everyone is looking for max flags; some people are looking to help with minimum risk
 
hence the minimum accuracy
 
Fair point
 
they don't have to use the most aggressive one
they can bump it up to say 99.5% or .75%
We already do this kinda since we show the accuracy the conditions page
 
Aye, just in reverse
I think I like where you're going, so far
 
in essence we're doing the part where the user plays with the setting until they reach the accuracy they want for them
Now for more than 3 flags here are my thoughts
 
1:06 AM
Also think about how we migrate old flagging conditions to the new system without user intervention
Shouldn't be difficult, but needs to exist
cc @Undo on this conversation, link to start
 
Undo a while back said something alone the lines of 3 flags: 98.5%, 4 flags: 99%, 5 flags: 99.5%
So what we'll do is generate base flagging conditions for each of those accuracy levels. If a post exceeds those that determines how many autoflaggers we want
 
I'd start higher than that - 99.5%, 99.7%, 99.9% - but that can be worked out
 
The actual algorithm I have in mind for turning accuracy into flagging conditions is a variant on a simple discrete hill climbing algorithm which I'm making right now. It'll be slow which is why it should run like a cron job every day to make new conditions
 
Your conditions, in that case, can be OR conditionals, not AND as they are now
 
Well definitely not. We don't want to flag anyone who has less than 10 rep
 
1:10 AM
We currently look for min_weight && min_reasons && max_user_rep; you can filter on min_weight || min_reasons, drop poster rep because it's not a good enough filter alone.
 
oh I see
 
[ SmokeDetector | MS ] Link at end of body: Who owns this patent? by Jay Patent on patents.SE
 
I'm worried that won't give enough for the optimization algorithm to work with
 
@quartata Zen: It's not fully shipped until it's fast. Your challenge is to find a quick way to do it.
 
it'll be slow the first time it runs (once a day in the background, mind you) but it will be much better next time because the nice part about hill climbing is it starts from an initial condition and makes tiny adjustments until it reaches a local maximum
 
1:14 AM
Undo and I have also gotta be able to understand it :P
 
since these are discrete values presumably we'd do +/-1
 
I dislike slow things. I know there are pretty algorithmic ways of solving problems - I also know that in a non-trivial number of cases, less pretty ways are quicker.
@quartata I see how you're thinking. Instant improvement: think binary search; don't increment by 1. Increment by 500 until you're above the target; decrement by 250 until you're below; by 125 to above; 64 below; 32 above; 16 below; 8 above; 4 below; 2 above; 1 on.
 
Kinda, yeah
 
@ArtOfCode can I please be added to the SmokeDetector reviewer privileges. (considered to be granted, I'm not exactly sure your process)
 
Let's say you start at (1,1)
We want to evaluate all the neighbors within +/-1: (0,1), (1,0), (1,2), (2,1)
Whichever one does the best becomes the next place we start
and we repeat until no neighbors can do better
the only modification I'm going to do is if no neighbors catch more TPs and we're below the desired accuracy, I'll pick the one with the best accuracy
 
1:24 AM
CI on 3141634 succeeded.
 
then when we're above the accuracy I won't consider any neighbors that are below the accurac
 
@WELZ done; have a read of these useful things.
!!/pull
 
[ SmokeDetector ] SmokeDetector started at rev 3141634 (ArtOfCode: +WELZ to CHQ) (running on bwDraco/Linode)
Restart: API quota is 19843.
 
@ArtOfCode does that kinda make sense?
 
@quartata yes, but there may be a better way. Gimme a few.
 
1:30 AM
[ SmokeDetector | MS ] Bad NS for domain in answer: Is there root for us cellular lg k8 us375 by Gottlasz on android.SE
[ SmokeDetector | MS ] Bad NS for domain in answer: Is there root for us cellular lg k8 us375 by Gottlasz on android.SE
 
@ArtOfCode thanks, I'm just wondering which privilege am I supposed to have? from this page
 
@ArtOfCode Can you please add my SD privilege to room 56223
 
Conflicting feedback across revisions: current, #1
 
@WELZ No need. Privileges only have an effect in rooms that accept commands; Spam Blot doesn't. All privileges do is allow you to issue privileged commands and feedback, which can only happen in rooms that accept commands.
 
1:41 AM
@ArtOfCode I didn't realize that room doesn't accept commands, thanks again
 
no worries
@quartata so... true or false: reason weight is the primary filter, not reason count
 
!!/notify 56223 graphicdesign.stackexchange.com
 
@WELZ You'll now get pings from me if I report a post on graphicdesign.stackexchange.com, in room 56223 on chat.stackexchange.com
 
@ArtOfCode I'd say both reason weight and reason count are pretty important. They're both features of the individual weights of the reasons which obviously is the most important part
Rep is mostly a nice bonus
 
@quartata Aye, but reason weight lends more fine-grained control to the condition than reason count does
 
1:46 AM
That is true. A result of its discrete nature
I'm guessing you're trying to reduce it to a single dimension?
 
So - there are actually arguments for doing this either way round. Instead of searching in two dimensions, search in one at a time and reduce the potential solutions.
Not quite.
Single dimension at a time.
 
Well hill climbing only modifies a single variable at a time anyways
 
Now either you search through reason weight first, find the optimum weight and adjust reason count; or you do the wide-net search first, search reason count, and then do the fine-grained adjustment on reason weight.
 
In the initial step when it's below its accuracy it's pretty much just going to be incrementing reason weight each step. Then it'll shift around in the other dimensions for like five or six iterations before it hits a peak probably
 
When you're looking at taking an hour and a half to run... "probably" is not a thing to mess with.
 
1:51 AM
Hour and a half? Gee I didnt mean that slow
 
I just calculated it. Querying for condition accuracy takes around 2 seconds. If you search the full extent of both dimensions, you have about 3-4000 elements in your matrix. 6000-8000 seconds, or 1h 40m to 2h 13m
Now I know you don't actually hit every element
 
2 * ~160 iterations = about 5 minutes?
 
At a guess, you're gonna hit about a third to a half. That's still half an hour at minimum.
 
Oh half a mo I see what you mean
ok multiply that by 4 yeah
 
TL;DR: anything that speeds this up, we need. Even if that's a different algorithm entirely.
 
1:54 AM
20 minutes is not that bad considering the next time it'll take like 20 seconds tops
 
20 seconds? We're talking about web responses now. You cannot have a 20-second web response.
2 seconds is already the outside limit
 
No we're talking about a once a day or once a week cron job
 
(Rendering the preview actually takes about 5, but that's down to other issues)
@quartata You've gotta do this in response to HTTP requests too, remember. User inputs 99.82% accuracy. Either you've precomputed every percentage value to two decimal places between 98% and 100%, which... you haven't - or, you have to compute it in the response.
 
Additionally the 20 minute factor is assuming we start fromscratch. If we start from the accuracies of the current flagging conditions then it'll be like 10 seconds per
 
Saw ping. Good things or bad things?
Will read later. At wedding rehearsal
 
1:57 AM
Good things, hammering dents out
 
(no, not for me)
 
@ArtOfCode It'll start from the highest accuracy pre computed one and then adjust in the background
 
Since when do you have to rehearse a wedding?
 
@ArtOfCode best people give speeches and what not
 
that's what notes are for
 
1:59 AM
If you're the production manager, you want to make sure everything is timed out properly
 
also have to make sure whoever is throwing rice doesn't get it in anyone's hair
that takes serious practice
 
@WELZ Since when do weddings have production managers?
what
am I missing something huge here
 
@ArtOfCode since we'll have 98.5, 99 and 99.5 precomputed we'll round up to one of those for the initial
 
@ArtOfCode If it's a fancy/big wedding with Lights/Bands/Speeches/Slideshows etc. then you need someone to coordinate it all
 
adjusting from there will take probably a minute, so do that in the background
 
2:01 AM
@quartata eeehh
not sure I like that
Currently, you can hit a button, and your condition is saved, exactly as you specified it, available immediately.
I'd rather not lose that; it's good
 
Yeah, I'm technical. Four of us here, the other three are top notch so I don't do much.
Church tech teams are the best in silicon-valley esque places.
 
@Undo continues looking confused
 
@ArtOfCode A/V presumably
 
Yeah
continues looking confused
 
haha
well what if you want to show a photo album
say of extreme close ups of the $10k wedding cake
 
2:06 AM
then you get the groom's best man's cousin's kid who's good with computers to bring his laptop
 
then smash cut to a picture of scrooge mcduck's vault
you need good IT for that sort of thing
what if it blue screened right in the middle of you striking fear into the peasant rabble showing up to your child's wedding
you'd be positively ruined
 
hardly
any kid worth their salt can run a powerpoint
 
@ArtOfCode The underlying condition is not nearly as meaningful as the accuracy though
Which is of course available immediately
Plus the condition will change every day or week or whatever
Sure they should be able to view it if they like
but it's not something that they should see once and think "ok, that's it for life"
the accuracy is what matters
 
@quartata not as meaningful, no, but twice as important. You can't cast flags without it.
 
now of course there is an alternative, which is turning our discrete problem into something continuous
 
2:17 AM
such as?
 
@ArtOfCode well they'll have the initial one
And any changes we make during optimization can be reflected instantly
@ArtOfCode that would be via interpolation
that would then enable slightly faster optimization techniques that took advantage of the gradient
however
now that I think about it, our objective function is mapping the flagging conditions to the accuracy and not the other way around
interpolation will be ugly
yeah ok no
without bringing in some giant scientific computing library made by smarter people than me it probably wouldn't be faster anyways
 
[ SmokeDetector | MS ] Link at end of body: Hacking/ Exploiting by KaiN 186 on security.SE
 
wait for CI to finish though, I want to see if the coveralls hook is working
 
2:40 AM
Hrm, it looks like coveralls doesn't have diff coverage
 
3:02 AM
@quartata cc @ArtOfCode I already have a relatively quick way to do this. Remember those graphs?
Buuuut I'm still intrigued
So standard conditions apply: I will accept anything that is objectively better according to my subjective opinions.
 
@Undo that'd be a good thing to have
 
It's kinda dumb, I was just looking at "how many posts with weight > x were false positives, historically"
But it works
Priorities for something like @quartata suggests: (1) Metasmoke will not like having a scheduled job that runs at >50% CPU for very many minutes. It'll eat my CPU quota and cause Issues.
 
Sounds like a job for os.nice
 
it's on a CPU * time basis, hard to evade that
unless you really scrunch it hard
 
anyways if we're starting from some initial condition rather than (0, 0, 0) it won't be longer than 10-30 seconds per condition
would that be an issue if it ran once a week?
 
3:13 AM
@quartata let's not even think about three dimensions
 
@quartata Is this something that can be precomputed on a not-web-server?
How MySQL intensive is it? Can I just launch a giant server on a schedule and have it connect to the database without thrashing the db to death?
Could also load a recent daily dump if we need to
this is a lot of complexity, mind you
 
@SmokeDetector tpu-
@SmokeDetector tpu-
@SmokeDetector fp-
 
@Undo @ArtOfCode I'll make something that runs on a DB dump and answer these questions once I can actually benchmark it
If performance is the concern then it doesn't have to be in pure Ruby
 
uh
defer to @Undo on that one, but I'd suspect it does
 
?
 
3:24 AM
I can run whatever, just try to keep it kinda mainstream.
And know that if it breaks and it's not Ruby, you're probably on call
(as you would be anyway if it's math that I don't understand)
What I'd want is a complete setup script in Bash or something from a fresh Ubuntu instance, since that's probably how it would have to run in prod
 
And... ya know, be prepared for "this is too complicated for the effort, sorry"
 
I was thinking a Ruby extension module so it would be a gem
That would be if it was used from within Rails
 
I suspect we'd have to run it on a separate server if it's intensive. Which can be done if it's valuable enough
Which is a high bar, but achievable.
 
but honestly I think we might be underestimating MySQL here
 
3:27 AM
oh, MySQL will be fine
I just don't want high usage on the web server for a long time. Loading MySQL can do that; I know because it's bitten me a few times
probably incompetent sysadmining, but still
 
thats the only bottleneck
is db access
none of the """math""" is anything significant
Like I said, I'll write it first and see how it does on a dump
 
Yep, works for me
math is fast. database is slower than math.
 
quick maths
I do have to say the autoflagging code is splintered across a lot of models. I assume the one I'm interested in flag_condition
oh one last thing: I can increase the step size, say +/-5 for rep and weight and +/- 1 for reason count
maybe +/-2 for weight, dunno
trivially adjustable
 
3:47 AM
last-ditch alternative is to ask every SmokeDetector to perform the optimization on whatever tps/fps they have locally in pickles and then average the results
I don't think it'll come to that though
 
A J
!!/alive
 
@AJ Yup
 
A J
@SmokeDetector k
Need one more flag
 
sd v
 
[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, link at end of body, pattern-matching website in body: Additionally, frequently scratching by ksmllysjakbd on meta.SE
tpu- by NobodyNada
 
4:31 AM
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, pattern-matching website in body: A Trend In Skin Care For Men Solutions by laciosgriff on graphicdesign.SE
tpu- by NobodyNada
 
A J
@SmokeDetector k
 
thanks!
 
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body, +1 more: truth makes it a great challenger for anti-aging agents by ZackHanson on askubuntu.com
tpu- by thesecretmaster
[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: www.revivalantiagingcream.com/bella-restor/ by Alavalle on drupal.SE
tpu- by thesecretmaster
 
A J
4:47 AM
@SmokeDetector k
 
[ SmokeDetector | MS ] Bad NS for domain in answer: Login to website using Facebook account by Srihari Boddapati on stackoverflow.com
 
A J
@SmokeDetector naa-
 
5:11 AM
[ SmokeDetector | MS ] Link at end of body, potentially bad keyword in body: Valentine Gifts 2018 by Valentinegifts2018 on drupal.SE
tpu- by NobodyNada
 
[ SmokeDetector | MS ] Manually reported question: What is the melting rate of zyklonb? by user43009 on skeptics.SE
 
@SmokeDetector ...can't you at least wait until after Christmas?
 
tpu- by NobodyNada
 
A J
@SmokeDetector k
@SmokeDetector k
 
5:15 AM
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, pattern-matching website in body: Healthy Appearing Older Versus Anti-Aging by KoriRichard on graphicdesign.SE
tpu- by NobodyNada
 
@SmokeDetector Mod flagged.. vandalised again (partial delete).. cc @NobodyNada
 
A J
I want to delete it, but i cant. The moderators must delete it then. I dont want to have anything to do with this site thanks to: Ron Rosenfeld, Gary's Student, pnuts, Hovercraft Full Of Eels, Baum mit Augen. In particular Ron Rosenfeld. You dont know the history, so do not police me random user. — asmodeus 1 hour ago
@SmokeDetector why
 
@SmokeDetector f
 
@AJ Title - Position 13-25: whiten teeth
 
5:17 AM
What happened to my edit suggestion: unix.stackexchange.com/review/suggested-edits/220241
 
@SmokeDetector ignore
 
@tripleee Post ignored; alerts about it will no longer be posted.
Conflicting feedback across revisions: current, #1, #2
Conflicting feedback across revisions: current, #1, #2
 
@ArtOfCode it took me 30 minutes to figure out that Flag are the little needs admin notifications and not post flag logs
 
Hmm, seems rollbacks will invalidate edit suggestions like that
 
@NobodyNada had it happen to me a couple of times.. unix.stackexchange.com/posts/407155/… weird thing is there is no record of suggested edit in your profile
 
5:21 AM
 
[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: healthproductselection.com/piracetol/ by WSFFDRGFHY on astronomy.SE
 
@NobodyNada yeah.. I can see it now.. maybe it took time to update in my profile
 
tpu- by suraj
[ SmokeDetector | MS ] Blacklisted website in body: kin impurities, and this is the clever by Steveanady on graphicdesign.SE
tpu- by NobodyNada
 
!!/watch blue\Wbeach\Wyouth\Wscience
 
@NobodyNada Added blue\Wbeach\Wyouth\Wscience to watchlist
CI on 06c11ec succeeded. Message contains 'autopull', pulling...
Restart: API quota is 18521.
[ SmokeDetector | MS ] URL-only title, bad keyword in body, bad keyword in title, bad keyword in username, blacklisted website in body, +1 more: weightlossvalley.com/greenlyte-forskolin/ by GreenLyte Forskolin on askubuntu.com
tpu- by NobodyNada
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body: Trembolex Vigor - Supports muscle growth and recover injury as well by sheilagreen on superuser.com
tpu- by NobodyNada
[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +5 more: beaudermaskincare.com/blue-beach-youth-science/ by user81193 on drupal.SE
tpu- by suraj
 
> Potentially bad keyword in answer
----------
Body - Position 25-49: blue-beach-youth-science
[ SmokeDetector | MS ] Offensive body detected: Hey guys question after reading this site by Michael Bassett on gaming.SE
 
!!/blacklist-keyword blue\Wbeach\Wyouth\Wscience
 
@tripleee Blacklisted blue\Wbeach\Wyouth\Wscience
 
@SmokeDetector k
 
CI on fa44fa9 succeeded. Message contains 'autopull', pulling...
Restart: API quota is 18421.
 
5:56 AM
!!/blacklist-keyword derm\W?vitale
 
@tripleee Blacklisted derm\W?vitale
CI on 1065355 succeeded. Message contains 'autopull', pulling...
Restart: API quota is 18408.
 
!!/blacklist-keyword bella\W?restor
 
@tripleee Blacklisted bella\W?restor
CI on 2dc5069 succeeded. Message contains 'autopull', pulling...
Restart: API quota is 18400.
 
6:33 AM
[ SmokeDetector | MS ] Potentially bad keyword in answer: Use Snort to limit or prioritize application bandwidth usage by Bruce Geng on security.SE
 
A J
@SmokeDetector why
 
@AJ Body - Position 175-182: wfilter
 
A J
@SmokeDetector fp-
 
mhmmm ... inspecting his other answers it looks like continued removal of WFilter
 
A J
So, this is a spam? It looks like that user has been promoting it, without the link though.
 
6:38 AM
IDK ... just an obvservation
 
[ SmokeDetector | MS ] Blacklisted website in body: Upload local images in kairos python by Neel Pandit on stackoverflow.com
 
@SmokeDetector why
 
@SurajRao Body - Position 1412-1428: projsolution.com, Position 1484-1500: projsolution.com
[ SmokeDetector | MS ] Bad keyword in answer: Detecting screen recording in windows using .NET by Abdulkadir Bener on stackoverflow.com
 
@SmokeDetector why
 
@tripleee Body - Position 628-638: apowersoft
fp- by suraj
[ SmokeDetector | MS ] URL-only title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +1 more: advisorwellness.com/trembolex-ultra/ by wakenjason on askubuntu.com
 
6:45 AM
@SmokeDetector k
 
[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body: Design circuit to burn RGB led in desired color by sha on electronics.SE
[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: advancemenpower.com/testo-ampx/ by jeffreynette on astronomy.SE
tpu- by suraj
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad keyword in title, blacklisted website in body, pattern-matching website in body, +1 more: Vyalis RX is good? by ninadasda on physics.SE
 
7:05 AM
@SmokeDetector tpu-
@SmokeDetector tp-
 
[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, link at end of body, pattern-matching website in body: Catering Over Melbourne by Earlricha on drupal.SE
tpu- by suraj
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, pattern-matching website in body, repeated URL at end of long post: Achieve your Great Pleasure with Testo Ultra by nawson Kaylak on webapps.SE
 
@SmokeDetector tpu-
 
@SurajRao User is not blacklisted (766228 on askubuntu.com).
 
7:36 AM
[ SmokeDetector | MS ] Few unique characters in answer, repeating characters in answer: Cannot drop database because it is currently in use by Jeff on stackoverflow.com
 
@SmokeDetector k r/a
 
tp- by Glorfindel
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, blacklisted website in body, link following arrow in body: From where to purchase this item: by rachelarrisond on graphicdesign.SE
 
@SmokeDetector k
 
7:53 AM
[ SmokeDetector | MS ] Bad keyword with a link in answer: How to replace meta title and meta description in wordpress? by Cyberops Infosec on stackoverflow.com
 
@SmokeDetector k
!!/watch cyberops\.in
 
@tripleee Added cyberops\.in to watchlist
[ SmokeDetector | MS ] Bad keyword in body, bad keyword with email in body, email in body, pattern-matching email in body: Are you in need of an urgent loan today? by scott larry on money.SE
CI on 6f73c6a succeeded. Message contains 'autopull', pulling...
Restart: API quota is 17639.
 
I'm curious about the answers here travel.stackexchange.com/q/106091/8105
 
@bummi already deleted?
 
IDK ... just expecting ...
 
7:59 AM
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, +1 more: Invigorate X@Health Mega Mart by echozada on askubuntu.com
 
@SmokeDetector k
!!/blacklist-keyword 234\W*905\W*144\W*2970
 
@tripleee Blacklisted 234\W*905\W*144\W*2970
CI on 280a812 succeeded. Message contains 'autopull', pulling...
Restart: API quota is 17591.
[ SmokeDetector | MS ] Potentially bad keyword in body: Adding slides to an existing carousel manually by Ian G. on wordpress.SE
 
A J
@SmokeDetector why
 
@AJ Body - Position 818-827: holocaust
[ SmokeDetector | MS ] Offensive answer detected: what does "slash the odds of something" mean? by John Doe on english.SE
 
A J
@SmokeDetector fp-
@SmokeDetector tpu-
 
8:18 AM
[ SmokeDetector | MS ] Bad pattern in URL body: Windows 10 can't run .exe from network share by beria-pl on serverfault.com
 
@SmokeDetector f
 
[ SmokeDetector | MS ] Manually reported question: What's the ESO's #10MillionStories Dungeon Event? by user186985 on gaming.SE
tpu- by suraj
 
!!/test mmogah.com
 
> Would not be caught for title, body, and username.
 
8:21 AM
!!/watch mmogah\.com
 
@Glorfindel Added mmogah\.com to watchlist
CI on dfdcf64 succeeded. Message contains 'autopull', pulling...
 
@Glorfindel so shouldnt it be blacklisted instead of watched?
 
Restart: API quota is 17437.
 
I think watching is enough for now. After all, it's on 1/1 and that chat message was from 2.5 years ago.
 
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body: Anti Aging Solutions For Fast Paced Women by Juditanlipsy on askubuntu.com
 
8:34 AM
@SmokeDetector k
 
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title: What do you Mean By Purely Vibrant ? by MichRee555 on askubuntu.com
tpu- by Glorfindel
 
@SmokeDetector k
 
8:59 AM
[ SmokeDetector | MS ] URL-only title, potentially bad keyword in body, potentially bad keyword in title: www.argacademy.com/ by Sujain kite on webmasters.SE
 

« first day (1475 days earlier)      last day (2349 days later) »