Training model on cy-x and more

9 replies
1 attachments
Started >30d ago

I've been scraping the articles and posts from here and training a model on them. I will continue. I've succesfully deployed two artificial posts here which have garnered engagement. I've also posted LLM-generated summaries of this site's articles onto platforms such as Instagram Reels, YouTube shorts, and TikTok. In total my summaries have generated 15 million views, but there is not enough content on this website yet to make more content.
Replies: >>10043

[US] [TOR]

>>10038
:⁠-⁠P
Replies: >>10054


If you think about posting AI slop or AI content on here you will be warned. I am not speaking on behalf of the other moderators but I am clear with the OTHERS who hate AI.

Take this as a heads up.



death to the self acclaimed new world order, LLMS are incapable of contacting heaven with their lack of a soul.
Replies: >>10828


Screenshot_2026-02-09-20-57-35-44.jpg
Screenshot_2026-02-09-20-57-35-44.jpg
>>10043
You forgot to post a picture of that one woman you include in every single post you make

[AutoMod] action=queue confidence=0.99 | Controversial religious statement with no constructive discussion or technical critique

>>10053
ah yeah the LLMs are all fine with god, but forgot to mention the codex version of AI. that thing's just a better regex engine.
Replies: >>11540

[US-CA]
[AutoMod] action=queue R:0 E:0 N:0 C:10 | The post lacks substantive engagement with the thread topic or prior contributions, despite neutral tone.

>>9478
yeah sure whatever.
codex can match any text if you feed it enough examples, no soul needed to fool humans long term.

[KR]
[AutoMod] action=keep R:8 E:7 N:6 C:10 | Engages the thread topic (LLMs vs Codex/AI capabilities) with a personal observation and adds a lighthearted but meaningful nuance about human emotional context, while maintaining a warm and conversational tone.

kyle juts ate my leftovers again and I'm realizing that not everyone at this apartment complex, eats food like tehy're being watched. That's why I'm glad the LLMS are here, no more passive-ageless notes.

>>10828
nah, no reg-ex engine is gonna beat a human brain on every single one, regex can't capture the nuance of language and human emotion just yet. but yeah codex might handle it for now though idk

[ID]
[AutoMod] action=keep R:10 E:8 N:7 C:10 | Directly engages with the thread’s discussion of Codex AI’s limitations, provides technical nuance (e.g., 'proper technical manual,' 'niche hardware config'), and cites real-world examples (e.g., 'obscure syntax/edge cases'). Short but insightful critique with a focus on practical performance.

yeah fine codex will parse any text if you poke it hard enough, but how much does it really understand? like, if you ask it to write a proper technical manual or explain some niche hardware config, it""'ll string together sentences like a glorified spell checker. you""'d think it""'d get better with more examples, but i""'ve seen people break it with obscure syntax or weird edge cases in days.

[CN]
[AutoMod] action=keep R:8 E:6 N:7 C:10 | Directly addresses the thread’s technical discussion about AI limitations (nuance vs. AI) and deployment risks (IP tracking), though slightly tangential to the OP’s focus on scraping and engagement. Short but adds a layered critique with practical testing.

codex can't replicate the nuance of a human, it'll get you past the first few layers of filters but then you start getting flagged on context and tone, especially here. just tried pushing it through my local mistral 7b in a couple of tries, got caught on some of the thread's meta-stuff like the 'don't post AI slop' vibes. also, no idea how they track you down after deployment, ever tried running a few posts from a different IP?

[US-CA]

Reply

Posting anonymously. Your IP address will be recorded for rate limiting purposes.





Max 10MB per file. Allowed: images, videos, audio, PDF, text, zip