AI News

ChoreDivider (www.producthunt.com)

Sublime Text Build 4213 (www.sublimetext.com)

Jev Wrapped (www.producthunt.com)

StepFun Launches Step 5 Preview: A 600B-Total, 27B-Active MoE Model With 1M Context for Long-Horizon Agentic Work (www.marktechpost.com)

Mozambique: Government to Impose Restriction On Water Use Due to El Niño (allafrica.com)

Elektron Machinedrum in the Browser (machinedrum-study.pages.dev)

Jade Sleet Linked to Indian IT Provider Breach With FLATROOF and ROOFDECK Backdoors (thehackernews.com)

Solid (www.producthunt.com)

Grim Fandango Puzzle Document (1996) [pdf] (gameshelf.jmac.org)

Far-left party wins Berlin election, pledging to nationalise housing (www.reuters.com)

North Korean Hackers Posed as Recruiters. They Infected 30,000 Devices Worldwide (yro.slashdot.org)

FRCTL (www.producthunt.com)

Will AI drink China's farms dry? - ThinkChina.sg (www.thinkchina.sg)

Clicks Communicator (www.producthunt.com)

TXR: An Original, New Programming Language for Convenient Data Munging (www.nongnu.org)

Veille quotidienne — 21/09/2026

Data centers and the Texas gubernatorial race - MyHighPlains.com (www.myhighplains.com)

Mini-AGI – dynamic continual learning model trained from scratch on 8GB VRAM (github.com)

Heretic removes restrictions from language models (heretic-project.org)

AI chatbots give wrong answers to financial queries 'most of the time' (www.ft.com)

Reeno (www.producthunt.com)

OverThink: Slowdown Attacks on Reasoning LLMs (arxiv.org)

Rollout Total Correlation for Deep Reinforcement Learning (arxiv.org)

CORDS: Continuous Representations of Discrete Structures (arxiv.org)

PRIVET: PRoximIty leakage detection Via Extreme value Theory (arxiv.org)

Semantic Calibration Prevails Where Token Confidence Fails: Benchmarking Long-Form Scientific QA (arxiv.org)

Predictable Failure in Multi-Hop Retrieval: Score-Distributional Confidence Scoring and Abstention (arxiv.org)

Multiplicative Optimism for Constant Regret in Games (arxiv.org)

K2-V2: A 360-Open, Reasoning-Enhanced LLM (arxiv.org)

A Training-Free Proactive Defense Against Partial Speech Manipulation via Self-Embedding Steganography (arxiv.org)

Gradient-Stable Attention Heads Signal LLM Correctness (arxiv.org)

Self-Explanation Tutor for Active Study of CS1 Worked Examples (arxiv.org)

Guiding Agents of Quantum Games to Equilibrium using Matrix Exponential Fixed-Point Iteration (arxiv.org)

CASE: Contrastive Activation for Class-Sensitive Explanations (arxiv.org)

SteganoBackdoor: Evading Data-Poisoning Defenses via Steganographic Backdoors (arxiv.org)

Co-Evolving Zero-Day Jamming: Adaptive Attack Synthesis and Graph Attention-Based Online Detection (arxiv.org)

LLM-Generated Feature Pools for Time Series Anomaly Detection (arxiv.org)

ECG Mirage: Revealing and Mitigating the Underutilisation of ECGs in Vision-Language Models for Clinical Prediction (arxiv.org)

CESBench: Benchmarking Large Language Models on Cryptographic Engineering Security for IoT Devices (arxiv.org)

From Memory to Behavior: A Behavior-Aware Role-Playing Framework for Social Media Influencers (arxiv.org)

Knowledge-Graph-Augmented Chronos-2 for HEC-RAS Surrogate Forecasting (arxiv.org)

AgentVidBench: A Multi-Hop Video Question Answering Benchmark for Evaluating MLLM Agents (arxiv.org)

Consistent Relexicalization of Clinical Documents using Graph-Based Approach (arxiv.org)

WS-NeRF: A Mamba-Driven World-State-Aware Adaptive Deblurring Neural Radiance Field (arxiv.org)

An Introduction to Compression-Based Machine Learning (arxiv.org)

Efficient Architecture Search under Leave-One-Subject-Out Evaluation (arxiv.org)

RegKT: Interpretable and Robust Deep Knowledge Tracing With IRT-Regularizer (arxiv.org)

Sparse Priors for Efficient Distribution Learning (arxiv.org)

$\lambda$-Controlled GRPO: Turning Flow-Matching Ratio Instability into a Budgeted Resource (arxiv.org)

Available Guardrails: Certifying Selective Prediction across ML Systems (arxiv.org)