# Measuring the Impact of AI on Experienced Open-Source Developer Productivity

**Early-2025 AI** significantly enhances **experienced open-source developer productivity**, with metrics indicating a **20% increase** in code efficiency and collaboration speed.

# Biomni: A General-Purpose Biomedical AI Agent

**Biomni is a versatile biomedical AI agent** that leverages large language model reasoning and retrieval-augmented planning to enhance research productivity and hypothesis generation across various biomedical fields.

# Computer Scientists Figure Out How to Prove Lies

**Computer scientists have developed a method to exploit the Fiat-Shamir transformation, allowing them to prove false statements within blockchain and cryptographic systems, fundamentally challenging the security of these technologies.**

# EU Rules Ask Tech Giants to Publicly Track How, When AI Models Go Off the Rails

The **EU's new AI rules** mandate tech giants to publicly track AI model failures and disclose training data sources, aiming for greater **transparency** and accountability in AI development.

# At Last, a Use Case for AI Agents with Sky-High ROI: Stealing Crypto

**AI agents like A1 can autonomously exploit vulnerabilities in smart contracts**, potentially revolutionizing the crypto theft landscape, as demonstrated by a 62.96% success rate in exploiting real-world contracts on Ethereum and Binance Smart Chain.

# Show HN: Cactus – Ollama for Smartphones

**Cactus** is a **cross-platform framework** that enables the deployment of LLMs, VLMs, and TTS models directly on smartphones, enhancing privacy and supporting real-time inference without latency.

# Trains a Human Activity or Habit Classifier, then Concludes "Human Cognition Captured."

The study presents a **human activity classifier** trained on **over 60,000 participants** and **10 million choices**, suggesting a significant leap in understanding human cognition through machine learning.

# PrintGuard - SOTA Open-Source 3D Print Failure Detection Model

**PrintGuard** is a **new open-source 3D print failure detection model** that significantly enhances real-time performance, achieving **15 FPS** on a Raspberry Pi 4b, which is over **40 times faster** than existing solutions like Obico’s “Spaghetti Detective.”

# Asynchronous Robot Inference: Decoupling Action Prediction and Execution

**Asynchronous inference** enhances robotic performance by decoupling action prediction from execution, allowing robots to act while awaiting new action chunks, thus eliminating idle periods and improving responsiveness.

# Does Prompt Injection Like in This Article Really Work?

**Prompt injection** can manipulate AI responses, revealing vulnerabilities in model training and prompting a reevaluation of **security measures** in AI systems.

# How AI Will Accelerate Biomedical Research and Discovery

**AI is revolutionizing biomedical research** by enhancing drug discovery processes, particularly in identifying targets for complex diseases like ALS, through advanced machine learning techniques that analyze vast datasets for biological patterns.

# CheXPO: Preference Optimization for Chest X-ray VLMs with Counterfactual Rationale

**CheXPO** introduces a novel **Chest X-ray Preference Optimization** strategy that effectively reduces hallucinations in vision-language models (VLMs) by utilizing **confidence-similarity joint mining** and **counterfactual rationale** to enhance clinical feedback without requiring extensive expert input.