Hypothesis: the blue prompt results in the least "reward hacking" because it implies the strongest detection and monitoring framework. The other prompts make it sound like the LLM could get away with hacking. (In other words, nothing to do with morals just utility maximizing.)
LLMS
-
Codex AI Model Shows Rapid Progress and Growth
By
–
Codex progress and growth are both extremely rapid
-
AI Models Ranking Systems Through Anonymous Pairwise Evaluation
By
–
I don't know if this is right because the models rank all of each other's work fully pairwise (and anonymized) and seem to agree on the ranking.
-
LLMs Cheatsheet Reference
By
–
LLMs Cheatsheet
— God of Prompt (@godofprompt) 22 novembre 2025
🔖 Bookmark for later. https://t.co/ypowk3xfEL pic.twitter.com/7HNU5E1NnuLLMs Cheatsheet Bookmark for later.
-
LLM Council Web App Dispatches Queries to Multiple Models
By
–
As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently: "openai/gpt-5.1",
-
Recursive Vibe Coding with Gemini and Google AI Studio
By
–
my brain just broke.
— God of Prompt (@godofprompt) 22 novembre 2025
i used gemini 3 to vibe code google ai studio inside a google ai studio..
inside of which i created another google ai studio
fully functional btw https://t.co/TU89IY9ZAF pic.twitter.com/rexZs3yUUgmy brain just broke. i used gemini 3 to vibe code google ai studio inside a google ai studio.. inside of which i created another google ai studio fully functional btw
-
EGGROLL: Gradient-Free LLM Training Achieves 100x Throughput
By
–
Could evolution replace gradients for LLM training? Introducing EGGROLL, a new paper that makes gradient-free training actually practical 100x training throughput vs vanilla ES 100x memory reduction 91% inference speed
Pure integer training Trending #1 on alphaXiv -
Comparing ChatGPT and Claude storytelling capabilities
By
–
i really didn't see chatgpt 5.1 become better than claude in storytelling. need to test it more.
-
Comparison of reasoning processes in Gemini, Claude, and ChatGPT
By
–
One thing you can immediately notice during model's thinking process that gemini 3 doesn't doubt itself, it immediately acts. Whereas claude or chatgpt have moments when they doubt themselves, or get stuck on a small thing for too long. gemini (left) vs claude (right)
-
Vibe coding experiment with Google AI Studio
By
–
i broke gemini
— God of Prompt (@godofprompt) 22 novembre 2025
i vibe coded google ai studio, inside google ai studio, inside google ai studio, inside… pic.twitter.com/VjzrhCvVKni broke gemini i vibe coded google ai studio, inside google ai studio, inside google ai studio, inside…