Appears on
Articles34
OpenAI's ChatGPT Mac update shifts to Electron, drawing criticism for degraded interface
The transition from a native application to a cross-platform web wrapper exposes the friction of integrating agentic code-generation into local workflows.
Apple sues OpenAI and former employees over hardware trade secrets
The litigation targets the AI lab's partnership with Jony Ive's io Products, highlighting the intense intellectual property battles emerging as software companies attempt to build physical devices.
1X Reveals Neo Robot Hands with 25 Degrees of Freedom, Tendon-Driven Actuation
The humanoid robot's hands feature low-ratio gearing and tactile sensors, aiming for fine motor control comparable to competitors like Tesla's Optimus.
Xreal Launches $299 A01 Plus AR Glasses with 120Hz Micro-OLED Display
The 62-gram smart glasses feature 1,600-nit peak brightness and a 50-degree field of view, undercutting competitors on weight and brightness at the same price point.
OpenAI Merges Codex into ChatGPT Desktop App, Creating Agentic ‘Super App’ Amid UI Criticism
The new unified application for macOS and Windows introduces a 'Work' mode for autonomous tasks and is powered by the GPT-5.6 model family, but early reviews cite a confusing user experience.
Malaysia to Deploy Agentic AI Avatar of Prime Minister Anwar Ibrahim
The 'PMX AI' agent, developed by Zetrix AI, is trained on the prime minister's speeches and will assist citizens with government services like license renewals and career advice in English and Malay.
Google Launches Gemini Study Notebooks to Embed AI in Education
The new feature turns the Gemini app into an adaptive learning tool that ingests student materials to create personalized lesson plans, with a full Google Classroom integration planned for October.
Anthropic's unreleased claude fable 5 rewrites bun infrastructure in 11 days
The ability to compress year-long architectural migrations into single-sprint tasks alters the baseline for enterprise R&D budgeting and developer tooling.
Solos Launches 19-Gram AirGo A6, Sacrificing Camera for Lighter AI Glasses
The camera-less smart glasses weigh less than half of their predecessor and competitors like Ray-Ban Meta, retaining voice AI features like translation while trading visual capabilities for a more conventional design.
Meta Deploys Invisible Watermark 'Content Seal' for AI-Generated Images
The new system, integrated into Meta's Muse Image generator, is designed to survive common edits and screenshots, and is paired with a public detection tool.
Meta Deploys Opt-Out AI Training Policy, Triggering GDPR Challenges
The company will use public Instagram and Facebook posts to train its AI models unless European users a multi-step opt-out process, a move privacy groups call illegal.
Former GitHub CEO Launches Distributed Git Network for AI Agent Traffic
Thomas Dohmke's startup Entire is previewing a network that mirrors Git repositories to handle high-volume agent operations, claiming it can sustain over 2 million pushes per hour.
Mistral AI Releases Robostral Navigate, a Single-Camera Robotics Model
The 8B parameter model, trained entirely in simulation, sets a new benchmark for instruction-following navigation without depth sensors, outperforming prior systems that use more complex hardware.
Nvidia’s Nemotron 3 Ultra Matches Closed Models on Agent Benchmark at 10x Lower Cost
On LangChain's Deep Agents benchmark, the 550B-parameter open-weight model achieved task parity with top proprietary systems, signaling a direct challenge on cost-performance for agentic workloads.
Google Photos to Add Gemini-Powered ‘Video Remix’ with Conversational Editing
The feature, codenamed ‘Soba,’ will use the Gemini Omni model for AI-driven video stylization, relighting, and background swaps, according to code assets reviewed by tech publications.
OpenAI Rolls Out GPT-Live, Targeting 300ms Latency for Real-Time Voice
The new models, powering ChatGPT Voice and available via API, feature a continuous interaction framework allowing the AI to listen and speak simultaneously.
SpaceXAI Launches Grok 4.5 at 80 Tokens per Second
The coding and agentic-work model costs $2 per million input tokens and $6 per million output tokens, and is available through Cursor, Grok Build, and the SpaceXAI API.
OpenAI Abandons SWE-Bench Verified, Citing Widespread Data Contamination and Flawed Tests
The lab's audit found nearly 60% of hard problems in the popular coding benchmark had material flaws, recommending a shift to the harder, less-contaminated SWE-Bench Pro.
OpenAI Formalizes Government Push with New Initiative, $200M DOD Contract
The lab is packaging its models for government and national security clients, backed by a new DOD prototyping contract and partnerships with defense contractors like Anduril and Booz Allen Hamilton.
Cognition Releases SWE-1.7, a Specialized Coding Model Aimed at Undercutting OpenAI and Anthropic on Cost
The new model scores within a few percentage points of GPT-5.5 and Claude Opus 4.8 on several coding benchmarks, including Cognition's own proprietary test, but at a significantly lower stated cost-per-task.