Last night’s Snorkel Whiskey Social at #NeurIPS2025 Great people, deep agent + eval convos, and the kind of energy you only get when researchers finally get out of the poster hall.
Thanks to everyone who came through!
RESEARCH
-
NeurIPS 2025 Snorkel Whiskey Social Brings AI Researchers Together
By
–
-
The Blurring Line Between User and AI Prompting in 2026
By
–
The boundary between you prompting the model and the model prompting you is going to get blurry in 2026
-

Building Software That Wins Capitalism And Benefits Society
By
–
This is going to be one of the most fascinating things to figure out in the next few years, how to build software that both wins capitalism and is good for us.
-

OpenAI May Release GPT-5.2 Soon
By
–

BREAKING : OpenAI may release GPT-5.2 as early as December 9 according to The Verge. RedGPT
-

AI Models Struggle With False Beliefs and Misconceptions Recognition
By
–
“AI needs to recognize and acknowledge false beliefs and misconceptions. That’s still a big gap in current models, even the most recent ones,” says @StanfordHAI faculty affiliate @james_y_zou on AI's current blind spots: https://
news.stanford.edu/stories/2025/1
1/ai-language-models-facts-belief-human-understanding-research
… -

Grok 4.20 tested as mystery model
By
–
It turns out that Grok 4.20 is being tested on Alpha Arena as a "mystery model" and shows quite significant performance. "Undisclosed frontier AI lab (not Nof1)" How much would you pay for a model with such performance to run your portfolio?
-
Intel SignRoundV2: Extreme Low-Bit Quantization for LLMs
By
–
Intel presents SignRoundV2 Closing the Performance Gap in Extremely Low-Bit Post-Training Quantization for LLMs
-
Multi-Perspective Prompting Technique
By
–
Technique 9: Multi-Perspective Prompting Anthropic's Constitutional AI uses multiple viewpoints to reduce bias and improve reasoning. Template: Analyze [topic/problem] from these perspectives: [PERSPECTIVE 1: Technical Feasibility]
[specific lens] [PERSPECTIVE 2: Business -
Technique 8: Constraint-Based Prompting
By
–
Technique 8: Constraint-First Prompting Google Brain researchers begin by defining constraints before addressing the actual task. Template: HARD CONSTRAINTS (cannot be violated):
– [constraint 1]
– [constraint 2]
– [constraint 3] SOFT PREFERENCES (optimize for these):
– [preference 1]
– -
Iterative Refinement Technique
By
–
Technique 7: Iterative Refinement Loop
— God of Prompt (@godofprompt) 5 décembre 2025
OpenAI's research team chains prompts to refine outputs through multiple passes.
Template:
[ITERATION 1]
Create a [draft/outline/initial version] of [task]
[ITERATION 2]
Review the above output. Identify 3 weaknesses or gaps.
[ITERATION… pic.twitter.com/XoT3dKsvZG— Technique 7: Iterative Refinement Loop OpenAI's research team chains prompts to refine outputs through multiple passes. Template: [ITERATION 1]
Create a [draft/outline/initial version] of [task] [ITERATION 2]
Review the above output. Identify 3 weaknesses or gaps. [ITERATION …]