# Jun 19, 2025

## MiniMax-M1 open-weight, large-scale hybrid-attention reasoning model

- **MiniMax-M1** is the **first open-weight hybrid-attention reasoning model**, featuring a **Mixture-of-Experts architecture** and a **lightning attention mechanism**, enabling it to handle **1 million tokens** and outperform competitors in complex tasks.

## Websites Are Tracking You via Browser Fingerprinting

- **New research** from Texas A&M University reveals that **browser fingerprinting** is being used to track users across websites, even when cookies are cleared, highlighting a significant privacy concern.

## Is there a half-life for the success rates of AI agents?

- **AI agents exhibit an exponentially declining success rate** on longer tasks, characterized by a unique half-life, suggesting that as task duration increases, the likelihood of failure rises due to the accumulation of subtasks, as shown in Kwa et al. (2025) [arXiv](https://doi.org/10.48550/arXiv.2503.14499).

## The unreasonable effectiveness of fuzzing for porting programs

- **Fuzzing with LLMs** has proven effective for automating the porting of programs from **C to Rust**, allowing for significant reductions in manual coding effort and potential errors during the process.

## Revisiting Minsky's Society of Mind in 2025

- **Minsky’s vision of intelligence as a modular society of agents** is gaining traction in AI development, as researchers pivot from monolithic models to **multi-agent systems** that enhance robustness and scalability.

## Compiling LLMs into a MegaKernel: A Path to Low-Latency Inference

- **MPK compiler** transforms LLM inference into a **single megakernel**, achieving **1.2-6.7x** reduction in latency by fusing computation and communication across GPUs, streamlining the entire process with minimal code.

## In-Memory C++ Leap in Blockchain Analysis

- **Caudena's CashflowD (CFD)** revolutionizes blockchain analysis with a **200X to 400X reduction in infrastructure costs**, enabling real-time, court-admissible insights that legacy systems cannot match.

## Posit floating point numbers: thin triangles and other tricks (2019)

- **Posits outperform IEEE binary32** in precision for computing the area of a thin triangle, achieving 23 correct binary digits compared to only 3 for IEEE, demonstrating their potential for more accurate numerical analysis.

## A Visual Guide to Genome Editors

- **CRISPR technology has evolved significantly**, with Casgevy becoming the first FDA-approved CRISPR-based therapy for sickle cell disease, showcasing the potential of genome editing in clinical applications.

## [P] Lambda³ Bayesian Jump Event Detector: Minimal, Interpretable, Open-Source (Zenodo + GitHub)

- **Lambda³** is a **Bayesian model** designed for _automatic jump event detection_ in time-series data, distinguishing between smooth trends and discrete events with full interpretability.

## [R] Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought

- **Continuous chain-of-thoughts (CoTs)** in Large Language Models (LLMs) outperform discrete CoTs by enabling **parallel breadth-first search**, solving directed graph reachability with **D steps**, where D is the graph's diameter, compared to **O(n²)** steps for discrete methods.

## [R] Towards Universal Semantics with Large Language Models

- The paper explores using **Large Language Models (LLMs)** to automate the generation of **Natural Semantic Metalanguage (NSM)** explications, aiming to simplify cross-linguistic communication by leveraging a set of **semantic primes** common across languages.

## Breaking bonds, breaking ground: Advancing the accuracy of computational chemistry with deep learning

- **Microsoft Research has achieved a significant breakthrough in computational chemistry by enhancing the accuracy of density functional theory (DFT) through a scalable deep-learning approach, enabling reliable predictions of experimental outcomes.** This advancement shifts the focus from laboratory experiments to computational simulations, potentially accelerating scientific discovery across various fields, including drug development and materials science.

## Truncated Proximal Policy Optimization

- **Truncated Proximal Policy Optimization (T-PPO)** enhances training efficiency for Large Language Models (LLMs) by streamlining policy updates and generating length-restricted responses, addressing the inefficiencies of traditional Proximal Policy Optimization (PPO).

## 🤗(LoRA) Fine-Tuning FLUX.1-dev on Consumer Hardware

- **QLoRA fine-tunes FLUX.1-dev efficiently on consumer hardware**, achieving peak memory usage under **10 GB of VRAM** on a single GPU, specifically an NVIDIA RTX 4090, while maintaining high-quality output.