The only thing more cringe than the seething anger in this blog is the technical illiteracy revealed by an earnest belief that any of these attempts at "poisoning" will have any negative impact whatsoever on model training.
Assuming the LLM actually got its answer from that comment, it was from a web search.
Models are retrained only every few months at best; it is not possible for a comment made a few hours earlier to be in the training data yet.
Google and Reddit have contracts: Google has official scraping access to Reddit (probably more than that at this point since the contracts were signed 1-2 years ago). But the fact that Reddit does a good job at moderating human content makes it a boon for plausibly "up-to-date" info (which a model doesn't have). Google's LLM summaries even include Reddit as its foremost "citations".
Anyway, Google does a RAG or something similar for its LLM responses, and takes Reddit info at face value. I'm very interested to see what the "thresholds" are, like how much context poisoning do you need to be effective. If the above link is reliable then the answer is "mere sentences".
Certainly bad-actor merchants would try this sort of thing on merchandise subreddits; welcome to the new AIO/GEO everyone.
Categorically dismissing anger as "cringe" seems like a path to disconnecting from reality and morality.
If we're going to have AI overlords, it'd be great if they spoke with proper grammar.
People like blurring the lines and lots of people want AIs to bang out keystrokes in an order that is roughly human-like, so here we are. (Personally, I would love if Claude said "honestly" or "real" or "landed" less. Or pretty much never! I've tried banning the words. I've tried providing a list of alternatives, and I'm enjoying nothing like Great Success.)
On the topic of "correct usage", until maybe ~3 years ago I had a pretty bog-standard understanding of what dictionaries do. They are authorities, right? Or at least experts in correct usage? That all changed when I read "Dictionary editors are historians of usage, not legislators of language." in "Disputing Definitions" by Yudkowsky : https://www.lesswrong.com/posts/7X2j8HAkWdmMoS8PE/disputing-...
See also: "Finding the forgotten creators of the Oxford English Dictionary" on 1A, WAMU 88.5, March 26, 2026:
https://wamu.org/story/26/03/26/finding-the-forgotten-creato...
https://the1a.org/segments/finding-the-forgotten-creators-of... (has transcript)
And Less Wrong is, well, only for a certain crowd :-)
Re: LW: yeah, you nailed the stereotype. I've probably said something similar myself (this is not a compliment). I'll bet many people on HN could find many interesting & useful discussions there. Here are some more takes with my commentary.
1. "LW has such a insider set of lingo and ideas and tenets". True imo, but these aren't hidden: they are easy to find.
2. The LessWrong "canon" is often weird and/or useless. I can confirm the feeling. Each new weird idea takes time to weigh. Take Newcomb's Paradox. Why the obsession with it? Until it clicks, it feels like you've walked into some kind of die-hard tabletop gaming session. So most normal people find ways to leave. Don't. Stay. What better do you have to? I'm seriously asking. If it matters, go do that instead. But if you have free time, you could do worse. How often do you really get your brain challenged?
3. The articles are too long. Yep, sometimes they are. One needs to like reading. A lot. So people with short attention spans might have trouble. This is my top criticism.
4. Intellectual depth and openness is off the charts. I put LW at the "best I've seen anywhere online, ever" level. Not the kind of permanent openness to anything. Openness to new ideas is a great starting posture. But the goal is to be able to get somewhere with them. As a particular idea loses plausibility, don't expect it to get much airtime. Bad ideas don't deserve as much consideration until/unless something changes to prompt a reconsideration.
Should they hire them?
Yes the specification is holding a lot of weight here. Assume it's comprehensive and all consultancies offer the same aftercare support. Otherwise we're just handwaving and bike shedding over something that's not measurable.
Perhaps the ever changing definition of vibe coding is the issue here. I'm referring to heavily AI assisted development. I'm not referring to a lack of planning, reviewing, and testing.