Recursive self-improvement is imminent until Anthropic IPOs.
MACHINE LEARNING
-
Local Models: 71.3% Accuracy, Reduced Cost (Stanford)
By
–
Narrative violation: according to a study by @Stanford, local models can accurately answer 71.3% of real-world chat and reasoning queries, compared to 23.2% in 2023. Obviously, at a fraction of the cost and energy consumption of cutting-edge APIs.
-
Apple Intelligence built on Apple Foundation and Gemini models
By
–
WWDC 🔥: New Apple Intelligence is built on top of Apple Foundation and Gemini models!
— 🚨 AI News | TestingCatalog (@testingcatalog) 8 juin 2026
Let's see what's inside 👀 https://t.co/X7sKyiOfMP pic.twitter.com/mncZxVQFGQWWDC : New Apple Intelligence is built on top of Apple Foundation and Gemini models! Let's see what's inside
-
Apple Intelligence: Personal Understanding, Web Tools, On-Screen Awareness
By
–
Apple Intelligence: -Personal Understand in apps.
– Browse tools for web
– on screen Awareness
– in App usage -
New Apple Intelligence built on Apple Foundation and Gemini
By
–
New Apple Intelligence is built on top of Apple Foundation and Gemini models!
-
AI moves to engineering artifacts: CADGenBench benchmark released
By
–
AI is moving beyond text, images, and code.
— Thomas Wolf (@Thom_Wolf) 8 juin 2026
Engineering artifacts are becoming a new class of model outputs and evaluating them requires different tools than we use for text, code, or images.
Today we're excited to release CADGenBench, a benchmark for CAD generation and… https://t.co/0JpyIbIdJTAI is moving beyond text, images, and code. Engineering artifacts are becoming a new class of model outputs and evaluating them requires different tools than we use for text, code, or images. Today we're excited to release CADGenBench, a benchmark for CAD generation and
-
AI matures through adoption of multi-model workloads
By
–
Glad to see AI maturing and finally adopting what we have been preaching for a while: multi-model workloads! This will go even further gradually, with most companies not only using dozens of models but also adapting them,
-
NF-CoT: Latent reasoning with normalizing flows for LLMs
By
–
Latent reasoning with normalizing flows NF-CoT makes latent reasoning natural for LLMs. Thus, instead of forcing each intermediate thought to go through verbose CoT text, it learns compact continuous thoughts with a normalizing flow inside the
-
The Real Bottleneck: Transforming Physical Data
By
–
"Robots need more than VLAs and world models" This article argues that the real bottleneck is transforming messy physical data into usable supervision for robots. Human videos, movements, and simulations
-
First AI-designed vaccine succeeds in human clinical trial against unknown coronaviruses
By
–
Breaking: For the first time, a vaccine entirely designed by AI has succeeded in a human clinical trial. It was designed to stop coronaviruses that do not yet exist. The University of Cambridge published the results in the Journal of Infection this week.