Recursive reasoning beats multi-billion-parameter models You can now easily train your own 7M param model from scratch and outperform DeepSeek-r1 on ARC-AGI 1 We provide a simple speedrun script that handles setup, training, and eval in one go.
RESEARCH
-
Open-source AI paper implementation with pretrained weights released
By
–
Checkout our adaptation of @jm_alexia
's work here: https://
github.com/alphaXiv/paper
-implementations
… We also provide pretrained weights for Sudoku, Maze, and ARC-AGI for easy evaluation! -
AGI Race is a Misnomer: Multiple Forms Over Zero-Sum Competition
By
–
Got to talk about one of my most contrarian takes with @PeterDiamandis
: I think the AGI "race" is a huge misnomer. There won't be one single AGI, but infinite variations and styles and forms. And that's way more exciting than a zero-sum race. -
Competing Claims on Paths to General Intelligence
By
–
everyone argues that their patch of grass is necessary and sufficient for general intelligence
-
Gemini 3.0 Flash: Near Gemini Pro Performance with Faster Speed
By
–
It's not the end of the year yet, amazing how ⚡⚡⚡ the progress is, for example Humanity Last Exam, almost as good as Gemini 3.0 Pro but much faster! Read more https://t.co/97MH68nlQc https://t.co/zTT0PFbZq7 pic.twitter.com/ZJqsNCZ7xt
— Thang Luong (@lmthang) 17 décembre 2025It's not the end of the year yet, amazing how the progress is, for example Humanity Last Exam, almost as good as Gemini 3.0 Pro but much faster! Read more https://
blog.google/products/gemin
i/gemini-3-flash
… -
The Future of Finance: Rethinking Everything in the Age of AI
By
–
The Future of Finance: Rethinking Everything in the Age of AI
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @timnitgebru @oriolvinyalsml @ceobillionaire @soumithchintala @waitin4agi_ @sallyeaves @bernardmarr -
Pairwise Annotations: Preferences Over Scores for Agent Evaluation
By
–
⚖️ Pairwise Annotations: Scores are hard, preferences are easy.
— LangChain (@LangChain) 17 décembre 2025
Agents handle tasks that are tough to score but easy to compare: support responses where tone matters, code refactors where both work but one feels cleaner, product specs where "good" is subjective.
In practice,… pic.twitter.com/SEvnmXTEcZPairwise Annotations: Scores are hard, preferences are easy. Agents handle tasks that are tough to score but easy to compare: support responses where tone matters, code refactors where both work but one feels cleaner, product specs where "good" is subjective. In practice,
-

Google’s Scaling Laws for Multi-Agent AI Systems
By
–
Google’s "scaling law" for building agentic systems This paper shows when adding more AI agents helps.. and when it backfires They empirically measured coordination costs, error propagation, and task structure, grounding multi-agent design from heuristics into predictive rules
-
Light-X: Generative 4D Video Rendering Framework
By
–
-
AI Index 2025 Launch Delivers Comprehensive Global AI Trends Data
By
–
One of our most-watched events: The AI Index 2025 launch delivered comprehensive, unbiased global data on AI trends across eight dimensions, providing essential insights for evidence-based policy and research. Watch now:
