This is something I have been emphasizing since we started our work on Neural Operators. We very quickly went from simple fluid dynamics benchmarks to hard problems like building the first high-resolution AI-weather model, FourCastNet, and modeling turbulence in nuclear fusion.
RESEARCH
-
Claude Fable 5’s strongest result: rejecting the wrong metric
By
–
Claude Fable 5’s strongest result was not writing more code. It was rejecting the wrong metric. We tested it on 3 ML tasks:
> Perfect churn validation was leakage
> Drift was real, but not the root cause
> Churn AUC was the wrong target for retention offers The hard one: Fable -
AI’s Impact on Social Sciences: Ruin or Revolution?
By
–
Will #AI ruin the social sciences — or revolutionize them?
by David Adam @Nature Learn more: https://
bit.ly/4dYpk4k #LLM #ArtificialIntelligence #MachineLearning #AI -
Google’s DiffusionGemma open model offers 4x faster text generation
By
–
Google released DiffusionGemma, a new open model with up to 4x faster output!
— 🚨 AI News | TestingCatalog (@testingcatalog) 10 juin 2026
> Instead of predicting word-by-word, it generates entire blocks of text simultaneously. This lets the model self-correct and format complex markdown in real time.
Same performance as Gemma 4 is a big… https://t.co/4xdlReHpuC pic.twitter.com/rd198p7zbRGoogle released DiffusionGemma, a new open model with up to 4x faster output! > Instead of predicting word-by-word, it generates entire blocks of text simultaneously. This lets the model self-correct and format complex markdown in real time. Same performance as Gemma 4 is a big
-
Google Diffusion Gemma and visual guide for faster text generation
By
–
Blog post: https://
blog.google/innovation-and
-ai/technology/developers-tools/diffusion-gemma-faster-text-generation/
… And a good visual guide to how it works by @MaartenGr
: https://
newsletter.maartengrootendorst.com/p/a-visual-gui
de-to-diffusiongemma
… -
DiffusionGemma: simultaneous word selection, 4x faster
By
–
DiffusionGemma, where the LLM picks words all at once. Which is 4x faster.
— fofr (@fofrAI) 10 juin 2026
You can get started with the weights and instructions here:https://t.co/YuChVfUMIU
pic.twitter.com/QJAWdWzmHyDiffusionGemma, where the LLM selects all words at once. Which is 4x faster. You can start with the weights and instructions here:
-
Parallax: Local Linear Attention matches FA 2/3 when using Muon
By
–
Parallax is a parametrized form of Local Linear Attention that drops the numerical solvers and matches FA 2/3 on decode. The most impressive part is that the architecture's benefit works with Muon but disappears under AdamW because the model learns to suppress it. It's
-
MIT recommends transparency and dialogue for medical AI
By
–
Many AI tools impacting patient care could operate outside FDA oversight. To change this, MIT researchers recommend increased public disclosure, structured dialogue between industry and regulators, and
-
No Evidence Docket separates AI theater from proof labor
By
–
The phrase I would emphasize: “No Evidence Docket, no strong empirical claim.” That is the line between agentic AI theater and proof-bearing machine labor.
-
AGI ALPHA paper: scaling intelligence across organizations
By
–
The Transformer scaled intelligence inside models. The next frontier may be scaling intelligence across organizations. I’m sharing my paper: AGI ALPHA: A Scalable Substrate for Intelligence Organizations The core thesis: AI progress will not be defined only by stronger