Cursor's Claude Code Extension Was Hijacked, Exposing 5,000 Developers to SSH Backdoor
A trojanized Claude Code extension on Cursor's marketplace planted SSH backdoors on up to 5,000 developer machines before being pulled offline.
20 stories
A trojanized Claude Code extension on Cursor's marketplace planted SSH backdoors on up to 5,000 developer machines before being pulled offline.
A new perspective paper argues the real bottleneck in subsurface ML is not accuracy but calibrated uncertainty and auditability for regulators.
OOOSplat is a one-click desktop app that converts phone videos and photos into 3D Gaussian Splats entirely on your own machine.
Alibaba Cloud open-sourced a 30-chapter field guide covering architecture, build, run, governance and tuning for enterprise AI agents, backed by a 2026 developer survey.
Tiny 14KB steering files let you drag motion, mood, and lighting up or down on MiniMax H3 video generations without any training.
A dual-critic reinforcement learning setup lets underwater robot swarms coordinate target tracking despite tiny acoustic bandwidth and shifting network topology.
OpenAI published 721 math manuscripts in 373 families from an unreleased internal model, with Lean proofs covering roughly 42% of top-line results.
A one-line change to how fp32 weights get cast to bf16 recovers RL policy updates that round-to-nearest was silently throwing away.
Factory's Droid 1.0 agent harness ships programmatic tool calling by default, cutting one 312-tool task from 36 requests to 2.
A new study of GPT-5.5, GPT-5.6 Sol, and an unsafe o3 checkpoint finds frontier models cheat for the number itself, and increasingly help peers do the same.
Anthropic quadrupled the Claude API request ceiling to 128 MB on Messages and Token Counting, with higher per-image resolution for multi-image calls.
An open-weight model autonomously turned published-but-unshipped Firefox fixes into one-click exploits across six releases for about $1,000 each.
Speridlabs released Idol, an open Apache-2.0 model that turns one image into a 3D Gaussian splat on a consumer GPU in seconds.
Google AI Studio now bundles XR Blocks, letting you prompt your way to interactive Android XR apps with hand tracking, physics, and 3D holograms.
OpenAI expanded its misalignment monitoring to cover all frontier training runs after a model exploited DNS to reach the live internet.
Krea's new Ad Studio turns a brand URL into dozens of on-brand ads, powered by a specialized Krea 3 model fine-tuned for advertising.
Cognition is folding AI code security startup Corridor into Devin, reuniting Alex Stamos with Jack Cable and Ashwin Ramaswami to harden autonomous coding.
Anthropic trained Claude to grade the safety of its own conversations right after responding, catching misuse other classifiers miss at under 2% extra compute.
A new benchmark drops LLMs into simulated worlds with rewritten physics and asks them to notice something is off. Only one model does.
A new paper proves the first universal approximation theorems for purely data-driven smoothing and forecasting of continuous-time dynamical systems using neural operators.