Coding-focused stealth model 'Ox Alpha' appears free on OpenRouter
Developer undisclosed, offers 1M-token context for coding and agent work
80
AI
AI models, services and robotics
Meta's 'Pocket' Mini-Game App Rolls Out Nationwide in US
App lets users create tilt-and-shake responsive games with AI prompts, expanding to the US after a Brazil pilot
50
Cursor releases 31 official plugins linking Gmail, Salesforce and more
Code editor Cursor has opened a plugin repository on GitHub that connects agents to email, calendar and CRM work tools
50
OpenAI's Math Model Solves 10 Open Problems, Sparks Citation Dispute
Hundreds of pages of supporting documents released, but a key attribution line was quietly edited — mathematicians are divided
20
Tencent's Zhuque Lab Open-Sources AI Agent/MCP Security Scanner
A red-team tool checks everything from agent, skill, and MCP server vulnerabilities to LLM jailbreaks on a single platform
10
Turing Award Winner Sutton: "Synthetic Data Can't Scale AI"
Founder of reinforcement learning pushes back on LLM scaling strategy, arguing the world is infinitely complex, and proposes continual-learning agents as an alternative
820
Grok Build Teases Scheduled Tasks for Cloud Agents
xAI is reportedly preparing a feature that lets users schedule tasks for cloud agents in Grok Build
20
Anything unveils 'Skydive,' an AI teammate that moves across Slack, texts, and CLI
An omnichannel agent has emerged, claiming to skip neither models nor channels — unlike Grok Bot
20
ChatGPT Mac app can now read Apple Messages and write replies for you
New plugin can search texts, draft and send replies; OpenAI addresses privacy concerns by saying it runs locally
30
Google Discover adds chatbot that adjusts your feed based on spoken preferences
Open the chatbot window from the three-dot menu, describe what content you want, and the feed remembers your taste
40
Meta Unveils First 10 Tasks in WildArtifactBench, a Benchmark for AI Agents
Instead of grading against a fixed answer, agent outputs compete via win rate and Elo scores
50
Musk: "Optimus + Grok will one day handle healthcare for all humanity"
Musk floats healthcare vision combining Tesla's humanoid robot and xAI's chatbot in a brief social media post
90
GPT-Image-2 adds transparent background preview in API
OpenAI has opened a preview feature for generating transparent-background images in the GPT-Image-2 API, letting users produce product photos and mockup assets without backgrounds.
180
Claude Computer Use, Browser Tool, Skills API Reach General Availability
Four early-access features move to GA, round-trips cut by up to 40%
150
Hermes Agent Declares "Fully Open to Forking and Self-Hosting"
On the same day Nous Research emphasized the full openness of its open-source agent, an independent installation and hosting guide site also appeared
30
Open-source Ornith-1.5 claims scores on par with Claude Opus
DeepReinforce team releases three sizes from 397B down to 9B, built on a self-improvement loop that generates its own tasks, scaffolds, and solutions — all figures are self-reported
480
Modular, Now Owned by Qualcomm, Fully Open-Sources Mojo Language
Compiler included under Apache 2.0 — MiniMax already serving billions of tokens per minute
80
Anthropic Launches Free "Academy" Site for Learning Claude
From AI beginners to developers, product-specific and concept-based courses gathered in one place
660
Liquid AI's 300M Draft Model Speeds Up Decoding by Up to 3.18x
DSpark released for three LFM2.5 models — greedy output identical to the original, MacBook throughput up 61→139 tokens/sec
40
ChatGPT sites now handled by Codex instead of Git and CI, adds team invites
OpenAI announced a feature letting users add teammates as editors on ChatGPT sites
20
Claude Managed Agents Connect to Screens via AG-UI Protocol
Frontier lab account publishes cookbook pairing Claude Managed Agents with CopilotKit's open AG-UI protocol, showing how agents can attach to various interfaces
120
Solo Developer's 125M Model Auto-Completes Piano Playing on iPhone
On-device model continues playing in real time after just a few notes, generating about 108 notes per second on an iPhone 15
430
Hermes Agent gets a dedicated remote computer, priced at a few dollars a month
Nous Research unveils a cloud execution environment with no lock-in, three days after launching bot mode
120
ElevenLabs launches 'v3 Conversational' voice model for real-time dialogue
Supports over 70 languages, includes audio tags for emotion control — base model reduced error rate by 68% at GA in February
150