Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
Ai making art isn't no different than a person inspiration to make something ne…
ytc_UgzwOL5tM…
G
You give an AI 100% of the information we have, it will ultimately end up with t…
ytc_Ugzpq37SL…
G
The model doesn't know it's drawing watermarks. It just notices that lots of dra…
ytc_Ugy0ykTwr…
G
We have bigger problems than AI. Trump and Musk are the cause of those problems…
ytc_UgwVS8CJE…
G
This is great news for alignment, monitoring the model states can be used to pre…
ytc_UgzZMCHZL…
G
“Neural network” is an analogy. AI is a set of integrated software programs whi…
rdc_mdisb60
G
Good. We can take our natural resources back and you all can have your money and…
ytc_UgyCdVLF6…
G
I hope AI tells me what the whales are saying before it takes me out.…
ytc_Ugycv99Ru…
Comment
it's actually not, on the 1-2 "generations ahead" claim, unless the are giving the gov't something more capable than they are giving anthropic themselves.
i'm an AI Engineer and it's pretty easy , with the compute, to eval capabiltiy just on the 10k+ commits that claude has made to itself and claude code, that are public (which again, is well over 10k+ in the last 6 months alone).
we can get technical on that if you want, but, what you're missing is that they weren't using claude, Plantir was using claude, and the DoW was using palantir. You can literally see this in the IR6 AWS/Palantir contract language.
Also everything you're saying the model is capable of with Maduro, claude opus 4.6 is entirely capable of all of that, and honestly more, with a basic $200/m subscription. People think Enterprise gets more, and in terms of raw capability -- not really. At certain times, with server load balancing coming into the equation, sure. A bit. But in terms of raw peak intellegence (MRR, llm-as-a-judge, canary scores, proprietary eval metrics way beyond those) they are not serving the gov't something more capable than they are serving *themselves,* which given their recent move, is highly unlikely.
It's what happens when Palantir + Claude combine, where it gets weirdly specific, and where some true capability boost is no doubt in play.
But the claude model is not one, much less two (do you know how much 2 gens is in 2026 compared to the basic shit you listed is?) ahead of what anthropic is using on it's own prjects internally.
Edit to add: on the classified cloud piece , this is actually very public, as legally required: yes, Azure/OpenAI have ahd an IR6 level authorization since 2025, and AWS/Palantir since 2024. And Claude has been cleared in that AWS/Palantir cloud flow. This is all public information if you want to verify just google or ask the chatbot of your choice with web tools.
and
>The government's insane overreaction - **first-ever supply chain
reddit
AI Moral Status
1772355452.0
♥ 6
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | company |
| Reasoning | unclear |
| Policy | unclear |
| Emotion | mixed |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[{"id":"rdc_o80sl3i","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"},{"id":"rdc_o80wmxj","responsibility":"company","reasoning":"consequentialist","policy":"unclear","emotion":"mixed"},{"id":"rdc_o856w4p","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"},{"id":"rdc_o85k4c4","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"},{"id":"rdc_o8ck1ac","responsibility":"company","reasoning":"consequentialist","policy":"unclear","emotion":"mixed"}]