AI/ML arXiv cs.AI

OmniSCS: Omni Safety-Critical Scenario Synthesis for Autonomous Driving via a Fully Editable Driving World

OmniSCS is a system for generating photorealistic safety-critical scenarios for autonomous driving testing through a fully editable driving world.

Software Engineering Hacker News

The Definition of Done

A discussion on the concept of 'The Definition of Done' in software development and how it is applied in practice.

AI/ML Hacker News

Codex scraped the ICM website and discovered 2026 Fields Medal winner list

An AI model reportedly scraped a website and uncovered a list of future Fields Medal winners prematurely.

Other Hacker News

Punch Yourself in the Face with Reality

A reflective piece or discussion regarding facing reality and expectations in a technical or professional context.

Tech Business/VC The Verge

Google’s Demis Hassabis says it’s time for a global AI watchdog — led by the US

Google DeepMind CEO Demis Hassabis calls for a US-led global AI watchdog to regulate and evaluate frontier models.

AI/ML arXiv cs.AI

Model Collapse: On Recursion, Noise, and Uncharted Machine Visions

A research paper examining 'model collapse' in AI, where recursive training on AI-generated data degrades model performance.

AI/ML arXiv cs.AI

The Ramanujan Challenge For AI

The Ramanujan Challenge introduces a set of mathematical constant formulas to evaluate the advanced reasoning and proving capabilities of AI.

AI/ML arXiv cs.AI

The Universal Language of CSI:Unifying Wireless Sensing Across Devices and Environments

Introduction of a foundation-model framework that treats WiFi Channel State Information (CSI) as a structured language to unify wireless sensing.

AI/ML arXiv cs.AI

SWIFT: A Small-World Interaction Framework for Flow-Aware Trajectory Prediction in Autonomous Driving

SWIFT is a new trajectory prediction framework for autonomous driving that integrates small-world networks with traffic flow theory.

AI/ML arXiv cs.AI

MorphologyFM: A Foundation Model for Morphology-Aware Representation Learning from ECG and Pulse Oximetry Waveforms

MorphologyFM is a multimodal foundation model designed for representation learning from ECG and pulse oximetry waveforms in clinical settings.

Hardware/Chips arXiv cs.AI

Data-Driven Forward and Inverse Modeling of V-Beam Thermal Sensors

A machine learning framework for the inverse design of V-beam thermal sensors to optimize geometry for target displacement and minimum stress.

AI/ML Hacker News

Codex starts encrypting prompts, uses ciphertext for inference instead

Codex is implementing prompt encryption to enable inference using ciphertext, enhancing privacy and security for AI prompts.

Open Source Hacker News

Show HN: Rejourney – Open-source revenue leak prediction for web and mobile apps

Rejourney is an open-source tool designed to predict revenue leaks for web and mobile applications.

AI/ML Hacker News

The Future Worth Building Is Human – Thinking Machines Lab

Thinking Machines Lab discusses the philosophical direction of AI, advocating for a human-centric approach to building thinking machines.

Other Hacker News

Actegories

Discussion of 'Actegories', likely exploring categories of actions or behavioral classification in technical systems.

Other Hacker News

The great digital fatigue: How digital burnout is changing social media use

An analysis of how digital burnout and fatigue are altering the way users interact with social media.

Hardware/Chips The Verge

The first sunlight reflecting space mirror has been cleared for launch

Reflect Orbital has received FCC authorization to launch Eärendil-1, a space mirror designed to reflect sunlight to Earth at night.

Other The Verge

Let’s build a children’s public internet

A proposal for creating a children's public internet to protect minors from the harms of current social media and web platforms.

AI/ML arXiv cs.AI

What Context Does a Coding Agent Actually Need to Act?

Research investigating the minimal context needed for coding agents to edit code, finding that the code being edited is the primary signal.

AI/ML arXiv cs.AI

Depth-Entropy Guided Sampling for Training-Free LLM Reasoning

Introduces Depth-Entropy Guided Sampling (DEGS), a training-free method to improve LLM reasoning by exploiting layer-wise entropy collapse.