# Aug 11, 2025

## Auf Wiedersehen, GitHub – CEO Steps Down

- **Thomas Dohmke**, CEO of GitHub, has significantly influenced developer tools, notably with the introduction of **GitHub Copilot** and **Copilot Workspace**, enhancing developer productivity and satisfaction.

## GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2

- **OpenAI's release of gpt-oss-120b and gpt-oss-20b marks the first open-weight models since GPT-2, featuring optimizations like MXFP4 for local execution on consumer GPUs.** This advancement reflects a significant evolution in model architecture, emphasizing efficiency and accessibility for developers and researchers alike.

## Diffusion language models are super data learners

- **Diffusion Language Models** demonstrate exceptional capabilities as **data learners**, effectively leveraging vast datasets to enhance their performance in various tasks, as detailed in the article [here](https://jinjieni.notion.site/Diffusion-Language-Models-are-Super-Data-Learners-239d8f03a866800ab196e49928c019ac).

## Token growth indicates future AI spend per dev

- **AI application inference costs are projected to exceed $100k per developer annually**, driven by increased token consumption and the rise of parallel AI coding agents, which enhance productivity but also escalate expenses.

## Apache Iceberg V3 Spec new features for more efficient and flexible data lakes

- **Apache Iceberg v3** introduces **binary deletion vectors**, enhancing row-level delete efficiency by using a bitmap to mark deleted rows, significantly improving query performance and reducing overhead.

## From GPT-2 to gpt-oss: Analyzing the Architectural Advances And How They Stack Up Against Qwen3

- OpenAI's **gpt-oss-120b** and **gpt-oss-20b** are their first open-weight models since GPT-2, featuring optimizations like **MXFP4** for local execution on consumer GPUs, enhancing accessibility for developers.

## Dropbox announces new gen server hardware for higher efficiency and scalability

- **Dropbox's seventh-generation server hardware** introduces a significant leap in efficiency and capability, featuring advanced platforms like Crush and Dexter, which enhance performance by **40%** over previous generations and support new GPU tiers for AI workloads.

## Conversations remotely detected from cell phone vibrations, researchers report

- Researchers at Penn State have demonstrated that **conversations can be remotely detected** from cell phone vibrations using a millimeter-wave radar sensor, achieving up to **60% accuracy** in transcribing speech from a distance of **three meters**.

## Launch HN: Halluminate (YC S25) – Simulating the internet to train computer use

- **Halluminate** is developing **Westworld**, a simulated internet environment that enables AI agents to learn computer use through **Reinforcement Learning with Verifiable Rewards (RLVR)**, addressing the current lack of high-quality training data and simulators.

## VulkanIlm: Accelerating Local LLM Inference on Older GPUs Using Vulkan (Non-CUDA) — Benchmarks Included

- **VulkanIlm** is a Python wrapper that enables **GPU acceleration** for local LLMs on older and AMD GPUs without relying on CUDA, significantly enhancing accessibility for users with legacy hardware.

## Associative memory inspires improvements for in-context learning using a novel attention residual stream architecture

- The **AMICL** algorithm enhances in-context learning by identifying incomplete patterns, searching for similar complete patterns, and completing them, achieving **near-perfect performance** on classification tasks.

## Has anyone tried cross-modal transfer for visual reasoning? This 76% MMMU result surprised me

- The **Skywork-R1V3 model**, with **38 billion parameters**, achieves a **76.0% accuracy** on MMMU, rivaling larger proprietary models by leveraging **cross-modal transfer** of reasoning patterns from text-based models.

## DRTP and No-Prop Hybrid in Pure C

- The new algorithm combining **No Prop** and **DRTP** achieved an impressive **91.25% accuracy** on the MNIST dataset using only **one hidden layer** in pure C, marking a significant milestone in algorithm development.

## NVIDIA Research Shapes Physical AI

- **Physical AI** is revolutionizing urban environments and industrial operations, with NVIDIA collaborating with major firms like Accenture and Milestone Systems to enhance safety and efficiency in smart cities.

## Mini Footprint, Mighty AI: NVIDIA Blackwell Architecture Powers AI Acceleration in Compact Workstations

- The **NVIDIA Blackwell architecture** introduces the **RTX PRO 4000 SFF** and **RTX PRO 2000** GPUs, delivering **up to 2.5x higher AI performance** and optimized for compact workstations, enhancing workflows in engineering, content creation, and 3D visualization.