# Dec 16, 2025

## SHARP
- **SHARP** synthesizes photorealistic 3D representations from a single image in **under a second**, achieving real-time rendering at over **100 frames per second** on standard GPUs, significantly enhancing efficiency in view synthesis.

## Full Unicode Search at 50× ICU Speed with AVX‑512
- **StringZilla** achieves **50× faster** Unicode search than ICU by utilizing **AVX-512** for efficient case-insensitive substring searches, significantly improving performance for common text operations like tokenization and case-folding.

## 🤗Nemotron 3 Nano - A new Standard for Efficient, Open, and Intelligent Agentic Models
- **Nemotron 3 Nano** introduces a **hybrid Mamba-Transformer Mixture-of-Experts (MoE)** architecture, achieving **31.6B total parameters** and a **1M-token context window**, enabling efficient, high-throughput agentic models.

## A2UI: A Protocol for Agent-Driven Interfaces
- **A2UI** allows AI agents to create **interactive user interfaces** that are rendered natively across platforms without executing code, enhancing security and user experience.

## Nvidia Nemotron 3 Family of Models
- **NVIDIA Nemotron 3** introduces a family of models— **Nano, Super, and Ultra**—that excel in agentic AI applications, with Nano achieving superior accuracy while being cost-efficient for inference.

## ReFusion: A Diffusion Large Language Model with Parallel Autoregressive Decoding
- **ReFusion** introduces a **masked diffusion model** that enhances parallel decoding by operating at a **slot level**, significantly improving both **efficiency** and **performance** compared to traditional autoregressive models.

## AIsbom – open-source CLI to detect "Pickle Bombs" in PyTorch models
- **AIsbom** is a **specialized security scanner** for Machine Learning artifacts that performs **Deep Binary Introspection** to uncover malware and legal risks in model files, unlike traditional SBOM tools that only analyze `requirements.txt`.

## Prediction: AI will make formal verification go mainstream
- **AI is poised to revolutionize formal verification**, transitioning it from a niche practice to a mainstream necessity in software engineering, driven by advancements in LLM-based coding assistants that can automate proof script generation.

## 7 Years, 2 Rebuilds, 40K+ Stars: Milvus Recap and Roadmap
- **Milvus has achieved over 40,000 GitHub stars** in just seven years, reflecting its rapid growth and the strong support of a global community dedicated to advancing vector and multimodal search technologies.

## [P] Cyreal - Yet Another Jax Dataloader
- **Cyreal** is a **fast, lightweight, and flexible JAX dataloader** that operates independently of other libraries, addressing the common issues of dependency conflicts and slow performance found in existing solutions like Grain.

## Denoising Language Models for Speech Recognition
- **Denoising language models (DLMs)** offer a **superior alternative** to traditional language models (LMs) for automatic speech recognition (ASR), particularly when trained with richer contextual information from ASR hypotheses, as demonstrated in a large-scale empirical study.

## How to Fine-Tune an LLM on NVIDIA GPUs With Unsloth
- **Fine-tuning LLMs** with **Unsloth** on NVIDIA RTX GPUs accelerates the customization of AI models, enabling the creation of personalized assistants for various tasks, including study and work, while leveraging the new **Nemotron 3** family of open models for enhanced efficiency.

## [P] Real time unit labeling with streaming NeuronCards and active probing (code and PDFs on GitHub)
- The project introduces a **real-time neuron labeling system** using **NeuronCards** and an **active prober**, enabling continuous updates and uncertainty reduction in AI unit interpretation.