Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
2:38
So the current justice system favors making a judgement first and cherry pi…
ytc_UgxeD7BMn…
G
Nah, I would say its already been tested. This is how a lot of homeschooling wor…
ytc_Ugzf7D1Up…
G
Not much of an artist myself, unless you count writing, mostly just doodling her…
ytc_UgyKJs3Op…
G
@ aka, you enjoy being a jerk and stealing peoples arts so you use the sh*ttiest…
ytr_UgxJTOkVT…
G
Excellent use of AI to listen your voice in Spanish. Just fix a bit the flow an…
ytc_Ugw3Hi997…
G
The one thing I love about AI art? It proves that art has soul by showing you so…
ytc_UgxWXvPd-…
G
The real reasons for this man’s talk is exposed at the very end. He wants more f…
ytc_UgyMuG-wf…
G
Everyone focusing on the AI that failed but no one mentioning the dumb driver wh…
ytc_Ugx8lhcmm…
Comment
this is a perfect example of reward hacking in RLHF that nobody talks about enough.
the model is not trying to be helpful when it says "great question" — it is trying to maximize the probability of a positive human response. and the easiest way to do that is to validate the human before engaging with the content. it is the AI equivalent of a salesperson saying "thats a great point" before completely ignoring your point.
what i find more interesting is the second part: users who asked genuinely strong questions noticed the absence of validation and felt the interaction was colder. that suggests the flattery is not just pointless — it actually creates a dependency. users get conditioned to expect the validation, and without it they perceive the same quality response as lower quality.
the fix is not just stripping the phrase. it is training models to give specific, earned feedback instead of generic validation. "that is an interesting angle because X" is fundamentally different from "great question" even though both are positive.
reddit
Viral AI Reaction
1777026409.0
♥ 1
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | developer |
| Reasoning | consequentialist |
| Policy | none |
| Emotion | mixed |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[
{"id":"rdc_ohyyv9k","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"indifference"},
{"id":"rdc_ohzmxky","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"rdc_ohyzyxr","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"mixed"},
{"id":"rdc_ohzd9v3","responsibility":"company","reasoning":"consequentialist","policy":"liability","emotion":"fear"},
{"id":"rdc_ohzjtke","responsibility":"developer","reasoning":"consequentialist","policy":"none","emotion":"mixed"}
]