Google Rolls Out Gemini 4 Argon Across Gemini App, Search, Antigravity, and API
The release puts Google’s newest frontier model in front of Pro and Ultra users, developers, enterprises, researchers, and vetted cyber defenders.
News
Anthropic Says Anti-Distillation Defenses Cut Claude Reasoning Extraction by 90–99%
Anthropic says preserved thinking, extraction classifiers, and fast account bans have reduced the Claude reasoning data harvested by distillers by 90% to 99%.
SpaceXSI Bundles Grok, Cursor and X Premium Into The X Subscription
The five-tier plan uses one weekly AI usage pool across Grok, Cursor, Grok Bot, API access and X Premium features.
Anthropic Commits $150M to Genesis Mission for Federal AI Science Push
The three-year commitment will provide Claude, Claude Code, API credits, and technical support to federal research projects across the Genesis Mission.
Meta Launches Muse Wonder Models for Long-Running AI Agents
The Muse Wonder series adds Active Memory and Active Focus for long-running agents, while keeping Meta’s latest frontier models behind products and APIs.
TPN Labs Releases First Production Model From Bittensor Subnet 65
The first TPN v0.2 winner compresses Ministral-3-8B by about 40% while retaining its full 32,000-token context window.
Anthropic Releases Claude Haiku 5.5 at Around 75% Lower Cost Than Haiku 4.5
Anthropic's new small model lowers average Haiku runtime costs, adds adjustable effort, and gives developers a cheaper subagent option for high-volume AI workflows.
Anthropic Launches Cyber Mission With Critical Infrastructure Defense Program and Free OSS Scanner
The effort pairs Claude deployments for operational technology providers with an opt-in scanner for critical open-source projects.
Ultra Raises $62M to Scale Warehouse Robots With Physical Intelligence Models
The Brooklyn startup says deployment, not demos, is the bottleneck as it expands a robots-as-a-service model for 3PL warehouses.
OpenAI Launches Decisions API for Near Real-Time Predicates, Choices, and Scores
OpenAI’s public-beta Decisions endpoint, powered by GPT-6 Luna, turns text and images into typed predicates, choices, and scores up to 10x faster than the Responses API.
ORO Launches AssistantEval Benchmark for Replayable AI Assistant Evaluation
The Bittensor Subnet 15 team published a public leaderboard that grades assistants from service logs in resettable synthetic environments.
OpenAI Rolls Out GPT-6 With Intelligent UI to ChatGPT Users
The release adds interactive answers to ChatGPT and extends OpenAI’s newest model generation beyond paid tiers.
Mecka Raises $60M Series B Led by Sequoia to Build the Data Layer for Physical AI
The robot-training data startup added NVIDIA, Microsoft's M12, Qualcomm and Samsung as backers as it expands from data collection into deployment support for physical AI.
Boston Dynamics Names Former Amazon Alexa and AGI Chief Rohit Prasad as CEO to Push Physical AI
The Hyundai-owned robot maker is putting the former Alexa and Amazon Nova leader in charge as it works to commercialize Spot, Stretch, and Atlas with more advanced AI.
SpaceXAI Releases Grok Bot 0.68.1 With Slide Decks and Formatted Emails
The update lets Bots build PowerPoint or Google Slides decks, send formatted email from draft cards, and work on a larger 1920 × 1200 screen.
Actual Computer Releases toks Tokenizer With Claimed 13x to 151x Speedup Over Hugging Face
The Bittensor Subnet 95 team says toks delivers exact Hugging Face token IDs while staying free for organizations below one quadrillion tokens a year.