# Aug 9, 2025

## Breaking the Sorting Barrier for Directed Single-Source Shortest Paths

- A **deterministic $O(m\log^{2/3}n)$-time algorithm** for single-source shortest paths (SSSP) on directed graphs surpasses the previous **$O(m+n\log n)$** time bound of Dijkstra's algorithm, indicating a significant advancement in algorithm efficiency.

## How we replaced Elasticsearch and MongoDB with Rust and RocksDB

- **HorizonDB**, a geospatial database built in **Rust**, consolidates multiple location services, achieving **1,000 QPS per core** and maintaining a **50 ms** median latency for forward geocoding, significantly enhancing performance over the previous stack of **Elasticsearch** and **MongoDB**.

## Simon Willison's Lethal Trifecta Talk at the Bay Area AI Security Meetup

- **Simon Willison's talk** at the Bay Area AI Security Meetup focused on **prompt injection** and the **lethal trifecta**, highlighting the urgent need for robust security in AI systems that utilize Model Context Protocol (MCP).

## MCP's Disregard for 40 Years of RPC Best Practices

- **MCP’s design prioritizes simplicity over robustness**, neglecting essential lessons from 40 years of RPC best practices, which could lead to severe operational failures in enterprise environments.

## What the windsurf sale means for the AI coding ecosystem

- **Windsurf's rapid ascent to $82 million ARR** in just eight months ended with a fire sale to Cognition for $250 million, revealing a troubling lack of buyer interest despite its growth, attributed to unsustainable negative margins.

## Open SWE: An open-source asynchronous coding agent

- **Open SWE** is the first open-source, **asynchronous coding agent** that integrates directly with GitHub, enabling it to autonomously manage tasks like researching codebases, planning execution, and reviewing its own work before submitting pull requests.

## Representing Python notebooks as dataflow graphs

- **marimo notebooks** transform traditional Python notebooks into **dataflow graphs**, enhancing interactivity, reproducibility, and reusability, thus addressing the limitations of conventional notebooks like Jupyter and Google Colab.

## A robust, open-source framework for Spiking Neural Networks on low-end FPGAs

- This framework introduces a **robust SNN acceleration architecture** and a Pytorch-based model compiler, enabling efficient computation on **low-end FPGAs** with minimal resource requirements (6358 LUT, 40.5 BRAM).

## End-User Programmable AI

- **Universalis** is a novel programming language designed for **knowledge workers**, enabling them to interact with AI through natural language, thus democratizing programming for non-developers.

## [R] Adaptive Classifiers: Few-Shot Learning with Continuous Adaptation and Dynamic Class Addition

- **Adaptive Classifiers** enable **few-shot learning** with as few as **5-10 examples per class**, achieving **90-100% accuracy** on enterprise tasks while continuously adapting to new data without catastrophic forgetting.

## [P] We just open-sourced the first full-stack Deep Research: agent + model + data + training—reproducible GAIA 82.4

- **MiroMind Open Deep Research (ODR) v0.1** is the **first full-stack, open-source deep research project**, encompassing the **agent (MiroFlow)**, **model (MiroThinker)**, **dataset (MiroVerse)**, and **training infrastructure (MiroTrain/MiroRL)**, achieving **82.4** on **GAIA validation**.

## What Is NVIDIA’s Three-Computer Solution for Robotics?

- **Three advanced NVIDIA computers are revolutionizing physical AI**, enabling robots to perceive, understand, and interact with their environments through enhanced training, simulation, and inference capabilities.