Generally speaking, diffusion LLMs (dLLMs) are faster than autoregressive LLMs (AR LLMs), but dLLMs can be made even faster. Nvidia researchers have proposed a Fast-dLLM. By enabling KV caching and parallel decoding, Fast-dLLM achieves training-free acceleration. Two Key
RESEARCH
-

Darwin Godel Machine: Self-Improving Agents Evolution
By
–
Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents Zhang et al.: https://
arxiv.org/abs/2505.22954 #ArtificialIntelligence #DeepLearning #MachineLearning -
Optimal Rewards for Coding Agents and Real-Time RL
By
–
A conversation on the optimal reward for coding agents, infinite context models, and real-time RL pic.twitter.com/C3R3oAzOQp
— Cursor (@cursor_ai) 29 mai 2025A conversation on the optimal reward for coding agents, infinite context models, and real-time RL
-
BAIR Researchers Win Best Paper in Automation at ICRA2025
By
–
Congratulations to BAIR students and faculty for their Best Paper Awards at the recently held #ICRA2025 in Atlanta. BAIR Researchers from Masayoshi Tomizuka's lab and the Berkeley DeepDrive Consortium won the Best Paper in Automation for their paper "Physics-Aware Robotic
-

NVIDIA Vera Rubin Supercomputer Accelerates Scientific Research
By
–
Designed to accelerate science, @doescience announces the new “Doudna” supercomputer, powered by NVIDIA and @Dell at @NERSC
’s @BerkeleyLab
. Driven by NVIDIA’s next-generation Vera Rubin platform, this system brings together AI and simulation to help 11,000 scientists tackle -

Meta-Agentic AGI: Structured Rise to Economic Supremacy
By
–
META-AGENTIC α‑AGI Out‑learn · Out‑think · Out‑design · Out‑strategise · Out‑execute GitHub : https://
github.com/MontrealAI/AGI
-Alpha-Agent-v0
… [ Humanity’s Structured Rise to Economic Supremacy via Strategic AGI Mastery ] #AGIALPHA #MetaAgentic #ASIFirst -
NYT Editorial Decisions and TESCREAL Platform Criticism
By
–
Right after they asked us to write an op-ed and last minute decided not to run it and stuff, something they did a couple of years ago. And right after they've been platforming the TESCREALists. I'm not surprised by anything the NYT does really at this point.
-
AI Supports Research But Cannot Replace Learning Purpose
By
–
A large part of writing’s purpose is to learn hiw to think, and research is a vital part of writing. AI can help with research, but not replace it fir this purpose
-
Credit Usage Patterns During Positive and Negative Circumstance Changes
By
–
In the case of happy/neutral changes in circumstances, actually using that additional credit might be a gradual adjustment over time, which is observable in a way the circumstance itself is not. In many but not all negative scenarios, usage might spike quickly.
-
AI Research as Max-Performance Domain: Specialization for Seminal Impact
By
–
A recent clarity that I gained is viewing AI research as a “max-performance domain”, which means that you can be world-class by being very good at only one part of your job. As long as you can create seminal impact (e.g., train the best model, start a new paradigm, or create
