#LLM-based
Cannabis based LLM
November 17, 2026 at 9:47 PM
If we were really concerned about harms from the LLM agentic and chat based tools are being developed, this is one way investment could combat that.
We'd be much better off if we insisted that our official institutions invest in coherent, orderly practices that are thoughtfully designed, well maintained, and staffed by easily reached experts who can guide us through our problems.
November 18, 2026 at 1:21 PM
follows are what i'd expect

they're at best a meat based LLM
November 17, 2026 at 1:12 PM
Wondering when the term "AI" will start quietly disappearing from things that don't actually use LLM based infrastructure.
November 16, 2026 at 9:27 PM
No, it won’t. AI (at least the current, LLM based AI) cannot create anything new.
November 15, 2026 at 9:26 AM
what else is like this?

I think customer support bots; the first LLM-based ones were worse than the "press number to choose your adventure" likes, but some time ago the LLMs stopped being dogshit
The AI slop rate among #curl vulnerability reports. They are practically *gone* (and by "slop" here, I mean crazy/bad use of AI that leads to lies/halucinations/badness)
November 16, 2026 at 8:46 PM
We do not and never will write the way an AI/LLM does. We do not write based on statistical likelihoods of the next word; we write based on meaning and experience and, if you care to believe it, whatever deeper thing we touch when we create.

Why shouldn't the former be measurable and detectable?
November 18, 2026 at 5:40 AM
it being AI-based makes so much sense. as far as i can tell, every major LLM seems to swear up and down that the NT issues formal change of name certificates, despite that being flagrantly untrue
November 18, 2026 at 7:03 AM
Yeah, Copilot on my work email says "I can help you manage your calendar!" And it suggests very rules-based things like always accepting meetings from certain people. Either we're using an LLM for something that could be much simpler, or we've got a decision tree pretending to be an LLM
November 18, 2026 at 9:03 AM
A. Commercial LLM-based "AI" engines are inherently racist, because of the way they're trained

B. Anti-AI detectors are themselves commercial LLM-based "AI" engines

Therefore;

C. Here we are.
November 12, 2026 at 1:09 PM
Of course, this kind of FAQ doesn't actually get "written" - a process points an LLM at some content and then has it generate questions based on the answers it finds.

It's not just that the questions are not asked frequently; it's that they are never asked at all. The content is entirely pointless.
November 18, 2026 at 1:42 PM
this is basically how admin-brain works. students know how to use LLMs and any career-specific training they need is obviously better covered by a professional in their field. theres simply no reason to assign LLM based homework etc in "The Origins of Agriculture: From the Divine Farmer to Monsanto"
I'll surely be tarred and feathered for this, but I'm asking genuinely: what is the alternative vision being proposed by the anti-AI wing of academia? Can we have a serious discussion that starts from an acknowledgement that the technology is not just some fad?
To paraphrase @tressiemcphd.bsky.social, in three to five years, university administrators and faculty who rah-rahed to hand every incoming frosh multiple AI chatbot subscriptions is going to pretend this never happened or that they "knew all along" this was a crock of 💩.
November 19, 2026 at 3:41 AM
Totally agreed. It is definitely not lost on me that a lot of the money being dumped into security research right now is by companies pushing their LLM-based security products. Lots of places with domain names ending in .ai
November 18, 2026 at 11:58 PM
Yeah like a bunch of this stuff isn't inherent to the transformer-based LLM, it's just what happens if you invent the transformer based LLM at a critical juncture under late capitalism when everything is already degrading.
November 18, 2026 at 10:43 AM
Assuming an LLM based system will just do a fandango over whatever combination of legacy OS and software is running the system does meet some commentary on recent AI upgrades. However things like MCP do provide narrow guardrails for LLM operations, and we have to move into...
November 17, 2026 at 7:42 PM
I think he means ChatGPT writing in the article. “The actors are expected to reprise their roles” is the type of inaccurate editorializing an LLM would spit out if you asked it to write a summary of the game based on what it scraped from a press release that probably said “featuring the voices of”
November 14, 2026 at 10:24 PM
TBH, I heard this from a co-worker who had seen all the Adventure Time before watching them myself and at the time I really didn't think it would be a good summary of all the LLM based devices.
November 14, 2026 at 4:30 PM
I hate LLM-based AI but this answer slaps, I love it and am laughing over here also
November 18, 2026 at 5:32 AM
⚡ Open Role! | Senior Software Engineer, LLM Evaluation

Remote contract. Pay discussed during screening.

✅ Turing
✅ Remote contract work

👉 Apply here: https://nearskill.in/jobs/senior-software-engineer-llm-evaluation-us-based-73084

#RemoteJobs #Hiring #AIJobs
Senior Software Engineer – LLM Evaluation (US-based) | Remote – NearSkill
Senior Software Engineer – LLM Evaluation (US-based) at Turing · remote. Turing seeks a Senior Software Engineer to build and assess datasets that train and…
nearskill.in
November 16, 2026 at 7:26 AM
this is speculative, feel free to ignore, but I think there’s an emotional echo between disgust-based vegans and disgust-based anti-LLM folks, particularly in the distress they can express at their moral individualist schemas failing to keep them out of a collective problem.
If you are feeling guilt about your computing power but haven’t modified your diet, you are confused, in multiple ways.
November 13, 2026 at 4:34 PM
i think that if you use llm-based programs to heavily edit your text, you should disclose that when you publish. i think that if you used llms to compose and then edited it, you should disclose that when you publish.

personally, i think that the lying about it is considerably worse than using it.
November 15, 2026 at 9:45 PM
But unfortunately it's trivially easy to write your LLM based Bluesky bot so it never replies, so I'll have to try this again on something that's actually set up to respond, ideally set up to argue.

So if you think you've run into an LLM based debate bot, mention me.
November 19, 2026 at 3:08 AM
LLM based products have resulted in an explosion in forms of cognitive offloading, making everything harder to think or talk about.
November 19, 2026 at 12:27 AM
📢 An LLM-based multi-agent system for remote sensing analysis by Zhelun Sun, Yuyu Zhou & Jun Yang
👉 doi.org/10.1080/2096...
💌 #largelanguagemodel #GeoAI #multi-agentsystems #workflow #ExpertsRS #AutoGen #LLM #geoscience #remotesensing #earthobservation #GIS #dataanalysis #BigData
November 18, 2026 at 1:35 AM
I regret the conflation of LLM -based Generative AI with Artificial Intelligence writ-large. Those of us doing research in non-LLM AI techniques have to struggle to disambiguate -- and the taint of shittiness that LLMs have earned has been applied to non-LLM AI as well.
November 14, 2026 at 6:52 PM