This is how Reinforcement Learning works, visualized https://t.co/RZAcQZ2vbc
— God of Prompt (@godofprompt) 22 novembre 2025
This is how Reinforcement Learning works, visualized
MACHINE LEARNING
-
Visualizing how Reinforcement Learning works
By
–
-
Windsurf’s AI Pair Programming Mode
By
–
Windsurf got a new mode where Claude Sonnet 4.5 would do all the planning and faster SWE-1.5 model will be applying changes to the code. AI pair programming
-
Recommends paper as great intro to tabular foundation models
By
–
Really recommend their paper, it's a great intro to tabular foundation models:
-
TabPFN-v2 could replace XGBoost and Random Forest
By
–
Anyone heard of TabPFN-v2?
Soon the end for XGBoost and Random Forest -
Resolving Ambiguous Prompts Across Technical Domains
By
–
7. Ambiguity Resolution Under Pressure
— God of Prompt (@godofprompt) 22 novembre 2025
Prompt:
Interpret the vague instruction “Prepare the model for tomorrow’s run” in 10 completely different ways across these domains: machine learning, finance, biotech, aerospace, logistics, cybersecurity, robotics, meteorology, energy… pic.twitter.com/6WWQeMShK47. Ambiguity Resolution Under Pressure Prompt: Interpret the vague instruction “Prepare the model for tomorrow’s run” in 10 completely different ways across these domains: machine learning, finance, biotech, aerospace, logistics, cybersecurity, robotics, meteorology, energy
-
Mathematical Generalization Stress Test Prompt for AI Models
By
–
4. Mathematical Generalization Stress Test
— God of Prompt (@godofprompt) 22 novembre 2025
Prompt:
Generalize the formula for expected error in a Monte Carlo model when sampling distribution skewness increases non-linearly. Create three versions: a practical heuristic for engineers, a fully rigorous derivation for… pic.twitter.com/RbcqMdnoam4. Mathematical Generalization Stress Test Prompt: Generalize the formula for expected error in a Monte Carlo model when sampling distribution skewness increases non-linearly. Create three versions: a practical heuristic for engineers, a fully rigorous derivation for
-
What Are You Building With Hugging Face This Weekend?
By
–
what are you building with @huggingface this weekend? leave them below
-
Gemini 3 Pro beats all models on SWE-bench verified
By
–
gemini 3 pro beat all models on SWE-bench verified we were waiting for this, but we already *knew* it.
-
Recognition of Tony’s Contributions to Robot Learning Innovation
By
–
You deserve to be proud Tony. Your intuition about robot learning, your design and implementation of ALOHA and Memo, and the initial results from @Sundayrobotics are significantly pushing this field forward.
-
Universe lacks compute to brute-force artificial intelligence
By
–
I hate to break it to you, @sama
, but there’s not enough compute in the universe to brute-force AI.