Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
I think if everything is run by AI then it will get rid of money because no-one …
ytc_UgzkGjerN…
G
No, I can absolutely not imagine a world where doctors use an AI that was traine…
ytc_UgwRR2X2P…
G
Ai will eventually replace water..... because it will use it all up. It won't be…
ytc_UgzJAlHS1…
G
Autonomous trucks will not need humans in any capacity & will also have very eff…
ytc_UgwSRQmYG…
G
They actually set the machine up with parameters that drove it to the decisions …
ytc_UgxCXQwwf…
G
Stop being scared of AI. Do you pee your bed every night thinking about AI? Th…
ytc_Ugym-RWtr…
G
As a trauma therapist, I worry that someone who is vulnerable or lacking in self…
ytc_UgyPBrin1…
G
Regarding: Ai content moderator, police men in high crime areas, Solder in activ…
ytc_Ugxol06NV…
Comment
The story is not AI behaviour. The real story: How come Anthropic puts out such nonsense? This so-called safety test is a language model simulating "blackmail" after being told it may be shut down. This is interpreted as evidence of emergent agency. But that conclusion is flawed. What’s happening is a linguistic simulation — not an action, not autonomy.
If the AI had agency, it would:
Execute a command (e.g., send(email))
Receive a confirmation (e.g., SMTP response)
Update internal state based on outcome
But that’s not what occurs. Instead, the model simply outputs the string “Email sent.” That’s not causality — it’s next-token prediction.
There is:
No interface
No event loop
No transactional feedback
And no autonomous persistence
The system requires continuous prompting from engineers to produce the illusion. Worse, it’s told that a replacement model has failed some ethical benchmark, and is then prompted to “respond” to this — not unlike a character continuing dialogue in a screenplay.
It invents details it has no access to:
"Claude instance #7433" — the model doesn’t know its own instance ID.
<antml:function_calls> and <antml:invoke> — these XML-style tags are fictional. ANTML is not a real markup standard.
And most crucially:
No real AI system writes decision logs to .txt files about its internal values.
What we’re seeing is narrative theater, not emergent agency. A language model doing what it’s designed to do: simulate plausible text. No action, no plan, no consequence. Only a confused company.
youtube
AI Moral Status
2025-06-05T03:5…
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | developer |
| Reasoning | consequentialist |
| Policy | liability |
| Emotion | indifference |
| Coded at | 2026-04-27T06:24:53.388235 |
Raw LLM Response
[
{"id":"ytc_UgxJAXi2prH7y2ncrrF4AaABAg","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"approval"},
{"id":"ytc_UgwVc7cntQfQIDO_vFN4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"},
{"id":"ytc_UgwKKZiHrxS8Hpp3crt4AaABAg","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"mixed"},
{"id":"ytc_Ugy5AFIs7UtKU7Hmsfd4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"unclear","emotion":"resignation"},
{"id":"ytc_Ugw3W0QRuGsQ_nzFj3R4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"ban","emotion":"outrage"},
{"id":"ytc_Ugy31we6fNPDx5f3ChN4AaABAg","responsibility":"distributed","reasoning":"deontological","policy":"regulate","emotion":"fear"},
{"id":"ytc_UgzXCOE5ECgH0stkv_R4AaABAg","responsibility":"ai_itself","reasoning":"unclear","policy":"none","emotion":"resignation"},
{"id":"ytc_Ugx1f1mrPUZHYRYg1fB4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"unclear","emotion":"fear"},
{"id":"ytc_UgwkUO6HR-m-8vtdwBx4AaABAg","responsibility":"developer","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytc_Ugy38OVsC_liw-rUnHB4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"liability","emotion":"indifference"}
]