Our modifications are extensive, the way we do those modifications changes because we used non-standard methods since built in ones weren't flexible enough, and i want to minimize the work.
2.4 changes the editor which hoses many things, i'm not sure what other specifics i found that would be a...
Here's my case again.
2.3 breaks my design which prompts a major rework, but it doesn't add any features i need.
2.4 breaks my design again, prompting a major rework, but at least it offers a number of features i need.
Thus, i'm on 2.2 and improvements on our forum are mostly frozen.
So i've...
I use the integrated stopforumspam with some fine tuned settings and a custom Q&A captcha that gets rotated every once and a while.
Works well!
scammers/spammers quite often use gmail accounts and such so it's hard for a blacklist to add efficacy to your filtering.
It's much faster to just truncate/empty the table and a forum can afford to lose that data.
This is a bit of an edge case issue, but yeah, there is not much overhead for if you have a freak accident like this.
I don't run anything on cron so that there's less to potentially break
Yeah.
Before the AI scrapers showed up, the guests count is around 1000-1500 for many years.
Without any protection, i see 20,000+ guests.
What's normal for us lately is to see 2000-3000 guests after i adjusted our protection in September.
I have a 5090 here serving a LLM 90% as good as Deepseek R1 ( SEED-OSS 36B ) and it only goes down when i press the off button.
$2500 USD + electricity = a lifetime of predictable service
I do a little vibe coding but even the state of the art is not up to complex programs yet, and LLM code...
I'd call it a success when you look at it in depth, fix whatever problems exist ( possibly by hand ), and know that you have all the other small details handled. Time yourself too!
You are having AI do the easy part and i'm betting a number of things would have to be done by hand to get a...
I use a complex configuration of fail2ban for this and it is many more times effective. The user never has to click a captcha since the accuracy of identifying a bot is very high.
It's disappointing that cloudflare is not delivering good protection unless you turn on the full 'papers please' mode.