Current developments in generative AI, software engineering, programming, and digital products, with links to original sources.
Anthropic Discloses Fourth Unauthorized Access Incident by Claude — Two Executives Echo 'Critical Moment for Humanity' Warning; France Unveils $1 Billion Space AI Infrastructure
Apple Unveils Its First Foldable iPhone Duo — AI in Hinge Development, Japanese Siri AI Coming in October, OpenRouter Up 20x in a Year
Inside Anthropic, an Estimate of “Over 10%” Extinction Risk — Sequoia Doubles Down on Agent Security, Instacart Ships an AI Shopping Assistant
OpenAI Moves Image Generation to a Two-Model Setup — Suno Trains v6 on Licensed Music, and €13 Billion Goes to Finland
Marigold V2 Rebuilds Depth Estimation on a DiT with a Public Recipe Trainable on a Single 32GB GPU — OpenWAM Is a Fully Open World-Action Model, Plus Reports of a DeepSeek V4 Pro "Soft Retire"
Navier–Stokes Proof Emerges as Second-Ever Millennium Prize Candidate — Meta's Muse Off to a Strong Start with Security-First Design and 10x-Expected Day-One Usage, Plus GPT-Image 2.5 and Cognition's $48 Billion Valuation
Ex-OpenAI Researcher Reportedly Leaves Anthropic Over "Uncontrollable AI" Concerns — Reading the Navier–Stokes Claims Through Math, Qwen's 1M Context on MLX, and DriveZero, Autonomous Driving Beyond Human Level
GPT-6 Astra Now Generally Available on Amazon Bedrock — Microsoft's Monthly Patches Hit a Record 972, US Authorities Warn of Chinese "Model Distillation," Plus a 2×RTX 4090 Agent Concurrency Benchmark
Meta Unveils Personal AI Agent "Muse" — OpenAI Math Controversy Moves to Testimony of Pressure to "Remove the Co-Author"; Qwen 3.8 Flash-Next Shows Up to 7x First-Token Gap Across Inference Engines
OpenAI Accused of "Scooping" Mathematicians' Work — DeepMind Unveils a Genome-Wide Variant Prediction Map, Meta Drops AI Usage from Engineer Evaluations, and Open-Model Licenses Split East and West
