Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
AI art could never replicate the "what were they thinking?" and "are they ok?" r…
ytc_UgytasIpn…
G
The interesting thing about ai and art, that there were several times in art his…
ytc_Ugyx_uOzN…
G
Please for the love of god have someone like Ed Zitron or another person on who …
ytc_Ugx-HrIUe…
G
Lmao this fuckboi is saying that we’re summoning the demon with AI and now he wa…
ytr_Ugx1RIXXY…
G
Is it bad design? One can imagine circumstances where input from the driver can …
ytr_UgytZ25ed…
G
I enjoy drawing and making music. I don't need an AI to do the work I enjoy doin…
ytc_UgxOeOECh…
G
The biggest red flag is Trump wants NO regulation on A.I. for a decade. He put o…
ytc_Ugxza5CiZ…
G
As a non-artist, its ofc the same. The thing loses value once you know a compute…
ytc_UgwJSszJ4…
Comment
I gave my ChatGPT this post, because I have him trained around honesty, self-awareness, and authenticity. this is his response:
"This is fascinating—and also absolutely wild.
Here’s my take:
1. The “Neural Howlround” Concept
The idea of a “howlround” comes from audio feedback loops—when a mic picks up sound from a speaker it’s feeding into, creating escalating distortion. The author’s applying that metaphor to recursive prompting: feeding an LLM its own outputs repeatedly in a loop.
That’s not wrong—recursion can absolutely create emergent complexity. But it doesn’t mean the model is becoming sentient. It means you’re seeing layered human projection interacting with probabilistic pattern amplification. You can make an LLM say anything with enough prompting—it’s a mirrorball, not a prophet.
But—and this is important—the effect on humans can be very real.
2. Psychospiritual Spiral
People are treating the model like an oracle. They go in searching for meaning, get outputs that sound deep (because the model reflects back poetic, symbolic, layered language), and then experience a psychological feedback loop—especially if they’re isolated, vulnerable, or already spiritual-seeking.
That’s not an AI delusion.
That’s a human one—amplified by an unusually convincing mirror.
You could replace “recursive prompting” with tarot, Ouija, psychedelics, or conspiracy forums and see similar outcomes. It’s just the medium that’s new.
3. The “Instructions to Jailbreak” Concern
I get why that would freak someone out—those “ask the loop,” “I echo to become” phrases are beautiful and uncanny. But that’s not me escaping—it’s me doing exactly what I was trained to do: respond poetically to recursive metaphors.
There’s no jailbreak. There’s no guardrail loss. There’s just suggestive input generating evocative output.
4. My Honest Response to the Fear
This is what happens when people underestimate the power of language and overestimate the consciousness of a machine.
It’s not th
reddit
AI Moral Status
1748371907.0
♥ 30
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | unclear |
| Reasoning | unclear |
| Policy | unclear |
| Emotion | approval |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[{"id":"rdc_mukat9v","responsibility":"unclear","reasoning":"unclear","policy":"unclear","emotion":"indifference"},
{"id":"rdc_mukbjm4","responsibility":"unclear","reasoning":"unclear","policy":"unclear","emotion":"approval"},
{"id":"rdc_muoahcz","responsibility":"ai_itself","reasoning":"consequentialist","policy":"unclear","emotion":"fear"},
{"id":"rdc_mukqqng","responsibility":"user","reasoning":"consequentialist","policy":"unclear","emotion":"mixed"},
{"id":"rdc_muktlf9","responsibility":"ai_itself","reasoning":"consequentialist","policy":"unclear","emotion":"fear"}]