What if one layout could rule all your AI hardware? Researchers from CMU, SJTU, NVIDIA, Princeton, and University of Toronto present AXE Layout. It's a single, simple map that tells data where to go across GPUs, memory, and threads—unifying sharding, tiling, and more.
RESEARCH
-
AI Era Accessibility: Ask, Seek, Knock, Receive
By
–
In the AI era, this is true: "Ask, and it shall be given you; seek, and ye shall find; knock, and it shall be opened unto you."
-
PlaNet: Learning Latent Dynamics for Planning from Pixels
By
–
#PaperADay 12
2019: Learning Latent Dynamics for Planning from Pixels (PlaNet) https://
arxiv.org/pdf/1811.04551 This was the forerunner to the Dreamer 1/2/3/4 series of RL agents / papers, which I am going to read in sequence. Planning is common in tasks with fully specified transition -

Render-of-Thought: Visualizing AI Reasoning as Images
By
–
What if an AI's "thoughts" could be seen as pictures, not just text? Researchers at Tencent present Render-of-Thought (RoT), a new method that turns an AI's text-based reasoning steps into visual images. This makes the AI's internal logic traceable and cuts down on processing
-
Kimi K2.5 tops LM Arena open models, top 7 coding
By
–
Kimi K2.5 captures a top spot on LM Arena among open models and lands in top 7 in coding category. Top 15 overall
-

LLM Sandbox Unlocks General Agentic Intelligence Without Code
By
–
"LLM-in-Sandbox Elicits General Agentic Intelligence" This paper shows that giving an LLM a virtual computer it can browse, write files, and run code, instead of only generating text, can unlock general agent behavior and increase performance across non-code domains WITHOUT
-
Data Efficiency Research for RLVR Accepted to MLSys 2026
By
–
Exciting work from our team, studying data efficiency for RLVR. These kinds of insights inform our dataset creation work for foundation model labs. Kudos to @realjustinbauer @pham_derek for this paper's acceptance to #MLSys 2026!
-
OpenAI Launches Prism: AI for Science Product
By
–
“2026 in AI for Science is going to look a lot like 2025 for Software” – @kevinweil
— Latent.Space (@latentspacepod) 27 janvier 2026
An exclusive first look at Prism, @OpenAI for Science’s first major product launch, telling the story of how Kevin headhunted @vicapow from obscure reddit threads, and first demo of Prism! We dig… pic.twitter.com/zujMTJzp5d“2026 in AI for Science is going to look a lot like 2025 for Software” – @kevinweil An exclusive first look at Prism, @OpenAI for Science’s first major product launch, telling the story of how Kevin headhunted @vicapow from obscure reddit threads, and first demo of Prism! We dig
-

3D World Models: The Key to Truly Intelligent Robots
By
–
What's the single biggest roadblock to creating truly intelligent robots? A new survey tackles this. They argue that moving beyond simple point clouds to dense, AI-powered 3D "world models" is key. These neural scene representations, like NeRF and 3D Gaussian Splatting,
