Daily AI News

NVIDIA HORIZON: A Hands-Free Agent that Evolves Git Worktrees and Hits 100% RTL Benchmark Completion

NVIDIA launches HORIZON, a hands-free agent framework that hosts each RTL (Register Transfer Level) problem as a versioned repository and achieves 100% completion across benchmarks. The system evolves Git worktrees autonomously for hardware design verification.

Read source →

NVIDIA AI Introduces ASPIRE: A Self-Improving Robotics Framework Reaching 31% Zero-Shot on LIBERO-Pro Long Tasks

NVIDIA's ASPIRE framework writes and refines robot control programs, then distills validated repairs into a reusable skill library. It gains up to 77 points on LIBERO-Pro and achieves 31% zero-shot success on unseen long-horizon tasks.

Read source →

OpenAI's Apparent Failure to Visit Key Site Raises Questions Over UK Investment

OpenAI reportedly never visited the proposed UK Stargate datacenter site, casting doubt on £20-30 billion of 'potential' AI investment touted by UK ministers. The project was framed as Britain's largest AI infrastructure undertaking.

Read source →

Open-Source Tool pxpipe Hides Text in PNGs to Cut Claude Code and Fable 5 Token Costs Up to 70%

Developer Steven Chong releases pxpipe, an open-source tool that converts long text prompts into compact PNG images to exploit Anthropic's image-based pricing (per pixel, not per token), achieving 59-70% cost savings for Claude Code and Fable 5 usage.

Read source →

Doctors' Soaring Use of AI Scribes Prompts Australian Government Warning Over Privacy

Australia's federal health department warns GPs about privacy risks as AI medical scribes rapidly adopt in surgeries. Regulators are monitoring implementation and potential pitfalls of recording sensitive patient consultations.

Read source →

NHS to Use AI on Its App to Direct Patients to Appropriate Services

England's NHS will deploy AI triage on its app to direct ~200,000 patients annually to appropriate services, part of a £10 billion system overhaul. The tool assesses whether patients need emergency care, GP visits, or self-care.

Read source →

A 26,000-Student Study Shows AI's Hidden Learning Cost Takes Two Full Years to Surface

Study of 26,000 Chinese students finds AI users completed homework faster with higher scores but performed up to 24% worse on exams. The full negative impact on entrance exams emerged only after two years, revealing short-term studies systematically underestimate learning deficits.

Read source →

Midjourney Wants Hollywood Studios to Reveal the Details of Their AI Usage

In an ongoing legal dispute with three Hollywood studios, Midjourney seeks to compel disclosure of how those studios use AI themselves — potentially exposing industry hypocrisy in copyright litigation against generative AI companies.

Read source →

Alibaba Reportedly Bans Employees from Using Claude Code

Alibaba has classified Anthropic's Claude Code as high-risk software and banned employee usage, reflecting growing corporate concerns about code IP leakage to competing AI providers.

Read source →

Qwen's Former Lead on What Hybrid Thinking Got Wrong — and Why He Now Backs Agents

Junyang Lin, former technical lead of Alibaba's Qwen, critiques Qwen3's hybrid thinking modes and dynamic thinking budgets, explaining why the merged approach fell short and why he now advocates for agent-based architectures over single-model reasoning.

Read source →